JP5503484B2 - Speech synthesis apparatus and speech synthesis program - Google Patents

Description

  The present invention relates to a speech synthesizer and a speech synthesis program for synthesizing speech corresponding to input contents of a character input device such as a keyboard.

  Conventionally, as a method for synthesizing speech corresponding to the input content of a character input device such as a keyboard, a method such as that proposed in Patent Document 1 has been used. In the speech synthesis method proposed in Patent Document 1, the waveform of a consonant phoneme and the waveform of a vowel phoneme are stored in advance in a memory, and when speech is synthesized, the consonant phoneme corresponding to the syllable to be synthesized from the memory is stored. Speech synthesis was performed by reading out the waveform and the waveform of the vowel phoneme and connecting them.

Patent Application Publication No. 59-52436

  However, when the speech synthesis method proposed in Patent Document 1 is used for an application that synthesizes and reproduces speech at the timing when a key of a character input device is pressed, there is an interval between pressing a consonant key and a vowel key. If this occurs, the waveform of the consonant phoneme and the waveform of the vowel phoneme cannot be properly connected, and even if the synthesized speech is played, there is a problem that it cannot be heard as a syllable that combines the consonant and the vowel. It was.

  Currently, many speech synthesizers are on the market, but speech synthesis is not performed unless the input completion button, which means that input has been completed, is pressed. For example, “a” is synthesized and reproduced. In addition to pressing the “A” key, it is necessary to press the input completion button. Therefore, it takes time until the voice is synthesized and reproduced, and there is a problem that the application cannot be appropriately used in an application that synthesizes and reproduces the voice when the key of the character input device is pressed.

  In addition, in a speech synthesizer that is currently available on the market, in order to synthesize muddy sound “ga”, it is necessary to press the “n” key after pressing the “ka” key. There is a problem that “ka” is reproduced first without being reproduced, that is, “ka” is erroneously reproduced.

  The present invention has been made in view of the above points, and by sequentially synthesizing speech corresponding to the input content of the character input device, the time required for speech synthesis is shortened and the accuracy of speech synthesis is improved. It is an object to provide a speech synthesizer and a speech synthesis program that can be used.

In order to solve the above-mentioned problem, a speech synthesizer according to claim 1 is a character comprising character keys each consisting of a kana character key, a dakuten key, a semi-dakuten key, an a lower case key, a ya line lower case key, and a ta line lower case key. A speech synthesizer that synthesizes speech corresponding to the input content of an input device, comprising a clear sound repelling discrimination means, a muddy sound discrimination means, a semi-turbid sound discrimination means, a stuttering discrimination means, a syllable discrimination means with a prompt sound, and a speech waveform A storage means, a voice synthesis means , a key code, a make code indicating that each key has been pressed, and a break code indicating that each key has been released from the character input device. , And a repetition signal output means and a repetition waveform storage means, and the character input device includes an accent lowering key, an accent increasing key, and an accent delimiter. In case further comprising a chromatography, a, comprises an accent signal output means, the sound waveform storage means, and a low sound waveform storage means, and a high sound waveform storage means, consisting of configuration and.

  According to such a configuration, the speech synthesizer detects a state in which the kana character key is pressed by the sound and sound repelling determination means, thereby obtaining a sound character or sound repellent character specified by the key code of the kana character key. Determine. Further, by detecting the state in which the kana character key and the muddy point key are pressed simultaneously by the muddy sound determining means, the muddy character specified by the key code of the kana character key and the muddy point key is determined. Further, by detecting the state in which the kana character key and the semi-voice point key are pressed simultaneously by the semi-voice sound determination means, the semi-voice character specified by the key code of the kana character key and the semi-voice point key is determined. Further, by detecting the state where the kana character key and the lower case key are simultaneously pressed or the state where the kana character key and the lower case key are simultaneously pressed by the stuttering discrimination means, the kana character key and A) The roaring character specified by the key code of the lower case key, or the key code of the kana character key and the lower case key. In addition, by detecting the state in which the kana character key and the tab row lowercase key are pressed at the same time by the syllable with syllable discrimination means, the syllable character with the prompt specified by the key code of the kana character key and the tab row lowercase key is detected. Determine.

  Then, the voice synthesizer, the muddy sound discrimination means, the muddy sound discrimination means, the muddy sound discrimination means, from the voice waveform storage means for storing in advance the speech waveform of each syllable consisting of clear sound, repellent sound, muddy sound, semi-turbid sound, stuttering and syllable with prompt sound. Speech waveform corresponding to each syllable character composed of a clear sound character, a sound-repelling character, a muddy character, a semi-voice character, a stuttering character, and a syllable character with a sound The speech is synthesized by extracting.

  In this way, the speech synthesizer constantly monitors the key input status of the character input device by each discrimination means. Then, when any key of the character input device is pressed by the operator (user) and a key code corresponding to the key is input by the operator (user), the speech synthesizer is syllable corresponding to the key code by each determination means. In addition to discriminating characters, the speech synthesis means sequentially synthesizes speech corresponding to the discriminated syllable characters.

At this time , the speech synthesizer monitors whether the make code and the break code are input from the character input device by the repeat signal output means, and generates a repeat start signal when the make code is input. When the break code is input, the repetitive end signal is generated and output to the speech synthesizer.

  Then, when the speech synthesizer extracts the speech waveform corresponding to each syllable character from the speech waveform storage means, and when the repeated start signal is input from the repeated signal output means, among the speech waveforms of each syllable, From a repetitive waveform storage means for storing a repetitive waveform indicating a waveform for a certain period in a section corresponding to a vowel, extracting a repetitive waveform corresponding to each syllable character, until a repetitive end signal is input from the repetitive signal output means, Speech is synthesized by repeatedly connecting repetitive waveforms in a repetitive section of a speech waveform.

  In this way, the speech synthesizer constantly monitors the input of the make code and break code from the character input device by the repeated signal output means, so that the operator is pressing any key of the character input device. Recognize whether it is present or separated. The voice synthesizer then operates after any key of the character input device is pressed by the operator (after the makeup code is input) until the key is released (until the break code is input). ) By repeatedly connecting the repetitive waveforms in the repetitive sections of the sound waveform by the sound synthesizing means, the sound with longer sound is synthesized only while the operator is pressing the key of the character input device.

Furthermore , the voice synthesizer inputs the key code of the accent lowering key, the key code of the accent rising key, or the key code of the accent delimiter key from the character input device by the accent signal output means before the key code of each character key. If the key code for the accent-decreasing key is input, an accent-decreasing signal is generated and output to the speech synthesizer. If the key code for the accent-rising key is input, An ascending signal is generated and output to the speech synthesizer. When a key code of an accent delimiter key is input, an accent delimiter signal is generated and output to the speech synthesizer.

  When the accent drop signal is input from the accent signal output means, the accent drop signal is received from the low speech waveform storage means for storing the low speech waveform indicating the low accent speech waveform of each syllable by the speech synthesis means. A voice is synthesized by extracting a low speech waveform corresponding to each syllable character determined by each determination means after being input. Further, when an accent increase signal is input from the accent signal output means, the accent increase signal is received from the high speech waveform storage means for storing the high speech waveform indicating the high accent speech waveform of each syllable by the speech synthesis means. A speech is synthesized by extracting a high speech waveform corresponding to each syllable character that has been input and determined by each determination means. Further, when an accent delimiter signal is input from the accent signal output means, the low sound corresponding to each syllable character of the first character determined by each determination means after the accent delimiter signal is input from the low sound storage means Extract a waveform and synthesize speech by extracting a high speech waveform corresponding to each syllable character after the second character discriminated by each discriminating unit after the accent separation signal is input from the high speech storage unit .

  In this way, the speech synthesizer constantly monitors the input status of the accent lowering key, accent increasing key or accent delimiter key of the character input device by the accent signal output means. When the operator presses the accent lowering key, accent rising key, or accent delimiter key of the character input device and the key code corresponding to the key is input, the voice synthesizing unit performs the key code. To extract a low speech waveform or a high speech waveform, and sequentially synthesize an accented speech.

Then, the speech synthesis program according to claim 2, the computer, and a program for functioning as a voice synthesizing apparatus according to claim 1.

According to the first and second aspects of the invention, since the speech corresponding to the input content of the character input device is sequentially synthesized, the input completion button provided in the conventional speech synthesizer becomes unnecessary, and the speech is synthesized. The time until playback can be shortened. According to the first and second aspects of the invention, only when a key code of a kana character key and a key code of a dakuten key are input simultaneously from a character input device, a syllable such as muddy sound is synthesized. Voices such as muddy sounds are not mistakenly synthesized and the accuracy of voice synthesis can be improved , and when the operator continues to press any key of the character input device, it corresponds to the key. Synthesize syllables with longer syllabes, which can increase the number of voice synthesis variations, and if the operator presses the accent down key, accent up key, or accent separator key on the character input device, the syllable accent Therefore, variations in speech synthesis can be further increased.

1 is a block diagram showing an overall configuration of a speech synthesizer according to a first embodiment of the present invention. It is the schematic which shows an example of the key arrangement | positioning of the character input device used for the speech synthesizer which concerns on 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the speech synthesizer which concerns on 1st Embodiment of this invention. It is a block diagram which shows the whole structure of the speech synthesizer which concerns on 2nd Embodiment of this invention. It is the schematic which shows an example of the speech synthesis by the speech synthesizer concerning 2nd Embodiment of this invention, (a) is a waveform diagram which shows an example of the speech waveform of "ka", (b) is a character input device. It is a wave form diagram which shows the result of the speech synthesis when the kana character key of "K" is pressed for 1000 msec. It is a flowchart which shows operation | movement of the speech synthesizer which concerns on 2nd Embodiment of this invention. It is a block diagram which shows the whole structure of the speech synthesizer which concerns on 3rd Embodiment of this invention. It is a block diagram which shows the whole structure of the speech synthesizer which concerns on 4th Embodiment of this invention. It is the schematic which shows an example of the key arrangement | positioning of the character input device used for the speech synthesizer which concerns on 4th Embodiment of this invention. It is a flowchart which shows operation | movement of the speech synthesizer which concerns on 4th Embodiment of this invention. It is a block diagram which shows the whole structure of the speech synthesizer which concerns on 5th Embodiment of this invention.

  A speech synthesizer and a speech synthesis program according to an embodiment of the present invention will be described with reference to the drawings. In the following description, the same configuration is given the same name and symbol, and detailed description is omitted.

[First Embodiment]
Hereinafter, the speech synthesizer 1 according to the first embodiment will be described in detail with reference to FIGS. The speech synthesizer 1 synthesizes speech corresponding to the input content of the character input device 2. Specifically, as shown in FIG. 1, the speech synthesizer 1 synthesizes speech corresponding to the input content of the character input device 2 in real time, and outputs the synthesized speech to the speech output device 3.

  Here, the character input device 2 inputs characters corresponding to the speech to be synthesized to the speech synthesizer 1. Here, as shown in FIG. 2, the character input device 2 includes a kana character key 2a, a dakuten key 2b, a semi-dakuten key 2c, a line (a line) small letter key 2d, a line (or line) small letter key 2e, and a key. The apparatus is assumed to be a keyboard or the like having a key arrangement of line (ta line) lowercase keys 2f. However, the key arrangement of the character input device 2 is not limited to that shown in FIG.

  The character input device 2 is one of the kana character key 2a, the dakuten key 2b, the semi-dakuten key 2c, the a lower case key 2d, the yaw lower case key 2e, and the ta row lower case key 2f, depending on the operator who wants to synthesize speech. When the key is pressed, the corresponding key code is output to the speech synthesizer 1. Here, the key code is a code for identifying the key of the character input device 2 pressed by the operator. For example, “A = 1”, “I = 2”, “U = 3”, etc. It is represented by numbers.

  The voice output device 3 reproduces the voice synthesized by the voice synthesizer 1. Here, the sound output device 3 is assumed to be a device such as a speaker capable of reproducing sound. The speech output device 3 receives the speech waveform of the synthesized speech from the speech synthesizer 1. Then, the voice output device 3 reproduces the synthesized voice by outputting the voice waveform.

  Here, as shown in FIG. 1, the speech synthesizer 1 includes a clear sound repelling discrimination means 11, a muddy sound discrimination means 12, a semi-turbid sound discrimination means 13, a stuttering sound discrimination means 14, and a syllable discrimination syllable discrimination means 15. I have. Hereinafter, each element which comprises the speech synthesizer 1 is demonstrated in detail.

  The clear sound and sound repellent determination means 11 is for determining a clear sound or sound repellent character from the key code input from the character input device 2. In other words, the clear sound repellent determination means 11 monitors whether or not the key code of the kana character key 2a (see FIG. 2) is input from the character input device 2, and detects the state where the kana character key 2a is pressed. As a result, the clear sound character or the sound-repellent character specified by the key code of the kana character key 2a is determined.

  Specifically, when the key code is input from the character input device 2, the clear sound / repellent determination unit 11 determines whether or not the number of the key code is one. Then, when the number of key codes is one, the clear sound and repelling determination means 11 refers to a correspondence table (not shown) in which the correspondence between the keys arranged in the character input device 2 and the key codes is described, It is determined whether the input key code corresponds to the kana character key 2a.

  Then, when the key code corresponds to the kana character key 2a, the clear sound / patter recognizing means 11 specifies the clear character or the sound repellent character indicated by the kana character key 2a using the correspondence table, and uses this as the syllable information. The result is output to the combining means 17. The correspondence table described above is stored in advance in storage means (not shown).

  On the other hand, when the number of key codes input from the character input device 2 is two or more, that is, when a plurality of key codes are simultaneously input, the clear sound / repellent discrimination means 11 ends the processing there. Further, the clear sound / repellent discrimination means 11 also ends the process when the key code corresponding to the kana character key 2a is not input from the character input device 2.

  As shown in FIG. 1, a key code is input from the character input device 2 to the clean sound / repeat sound discrimination means 11. Then, the clear sound and sound recognizing means 11 generates the syllable information of the clear sound or the sound repellent character by the above-described method and outputs it to the speech synthesis means 17.

  The muddy sound discriminating means 12 discriminates muddy sound characters from the key code input from the character input device 2. That is, the muddy sound discrimination means 12 monitors whether or not the key code of the kana character key 2a (see FIG. 2) and the key code of the muddy point key 2b are input from the character input device 2, and the key code of the kana character key 2a. And the key code of the dakuten key 2b are detected at the same time, the muddy character specified by these key codes is discriminated.

  Specifically, when the key code is input from the character input device 2, the muddy sound discrimination means 12 determines whether the number of the key code is two and whether the two key codes are input simultaneously. Determine whether or not. Then, when the number of key codes is two and these are simultaneously input, the muddy sound discrimination means 12 refers to the correspondence table (not shown), and the two key codes input are It is determined whether or not it corresponds to the kana character key 2a and the dakuten key 2b. When the key code corresponds to the kana character key 2a and the dakuten key 2b whose key code is “Kara line, Sa line, Ta line, Ha line”, the muddy sound discrimination means 12 identifies the muddy character indicated by these keys, Is output to the speech synthesizer 17 as syllable information.

  For example, when the key code corresponding to the “Hi” kana character key 2 a and the key code corresponding to the cloud key 2 b are simultaneously input from the character input device 2, the muddy sound discrimination means 12 Is specified and output to the speech synthesizer 17 as syllable information.

  On the other hand, the muddy sound discriminating means 12 is used when only one key code or three or more key codes are input from the character input device 2, or when two key codes are not input simultaneously from the character input device 2. Then, the process ends. The muddy sound discriminating means 12 performs the processing even when the key code corresponding to the kana character key 2a or the muddy key 2b is not inputted from the character input device 2. Exit.

  As shown in FIG. 1, a key code is input from the character input device 2 to the muddy sound discrimination means 12. Then, the muddy sound discriminating means 12 generates syllable information of muddy character by the above-described method, and outputs it to the voice synthesizing means 17.

  The semi-turbid sound discrimination means 13 discriminates a semi-voice sound character from the key code input from the character input device 2. That is, the semi-turbid sound discrimination means 13 monitors whether or not the key code of the kana character key 2a (see FIG. 2) and the key code of the semi-turbid key 2c are input from the character input device 2, and the kana character key 2a. By detecting a state in which the key code of the semi-voice point key 2c and the key code of the semi-voice point key 2c are pressed at the same time, the semi-voice character specified by these key codes is discriminated.

  More specifically, when the key code is input from the character input device 2, the semi-turbid sound discrimination means 13 determines whether or not the number of the key code is two, and the two key codes are input simultaneously. It is determined whether or not. When the number of key codes is two and these are input simultaneously, the semi-turbid sound discrimination means 13 refers to the correspondence table (not shown), and the two key codes input are It is determined whether or not the key corresponds to the kana character key 2a and the semi-voice point key 2c. When the key code corresponds to the kana character key 2a and the semi-turbid point key 2c, the semi-turbid sound discriminating means 13 specifies the semi-turbid sound character indicated by these keys, and synthesizes the voice as the syllable information. Output to means 17.

  For example, when the key code corresponding to the “hi” kana character key 2 a and the key code corresponding to the semi-voice point key 2 c are simultaneously input from the character input device 2, "Pi" is identified and output to the speech synthesizer 17 as syllable information.

  On the other hand, the semi-turbid sound discrimination means 13 has only one key code or three or more key codes inputted from the character input device 2, or two key codes are not inputted simultaneously from the character input device 2. If so, the process ends there. Moreover, the semi-turbid sound discrimination means 13 also ends the process even when the key code corresponding to the kana character key 2a or the semi-turbid sound key 2c of "ha line" is not input from the character input device 2.

  As shown in FIG. 1, a key code is input to the semi-turbid sound discrimination means 13 from the character input device 2. Then, the semi-turbid sound discrimination means 13 generates syllable information of the semi-turbid sound character by the above-described method, and outputs it to the speech synthesis means 17.

  The stuttering discrimination means 14 discriminates stuttering characters from the key code input from the character input device 2. That is, the stuttering discrimination means 14 monitors whether or not at least the key code of the kana character key 2a (see FIG. 2) and the key code of the lower case key 2d are input from the character input device, and the kana character key 2a. By detecting a state in which the key code and the row-lower-case key 2d are simultaneously pressed, the stuttering character specified by these key codes is determined. The stuttering discrimination means 14 monitors whether or not at least the key code of the kana character key 2a and the key code of the yaw small letter key 2e are input from the character input device, and the key code of the kana character key 2a. By detecting a state in which the YA row lowercase key 2e is simultaneously pressed, the stuttering character specified by these key codes is determined.

  Specifically, when the key code is input from the character input device 2, the stuttering determination unit 14 determines whether or not the number of the key code is two or more, and the two or more key codes are simultaneously received. It is determined whether or not it has been input. Then, when the number of key codes is two or more and these are simultaneously input, the roaring determination means 14 refers to the correspondence table (not shown), and the two or more input key codes are Whether or not it corresponds to the kana character key 2a and the a-line lower-case key 2d of at least "ka line, sa line, ta line, na line, ha line, ma line, la line", or at least "ka line" It is determined whether or not it corresponds to the kana character key 2a and the yaw lowercase key 2e. The stuttering discrimination means 14 corresponds to the kana character key 2a and the a-line lowercase key 2d whose key code is at least “ka line, sa line, ta line, na line, ha line, ma line, la line”, Alternatively, if at least “K”, “S”, “T”, “N”, “H”, “M”, “R” are corresponding to the Kana character key 2a and the Ya-line lowercase key 2e, the roaring characters indicated by these keys are specified. This is output to the speech synthesizer 17 as syllable information.

  The stuttering discrimination means 14, for example, when the key code corresponding to the “F” kana character key 2 a and the key code corresponding to the “A” lowercase key 2 d are simultaneously input from the character input device 2. Then, the “F” of the roaring character is specified, and this is output to the speech synthesizer 17 as syllable information. The stuttering discrimination means 14 receives, for example, a key code corresponding to the “hi” kana character key 2 a and a key code corresponding to the “ya” lowercase key 2 e from the character input device 2 at the same time. In the case, the stuttering character “Hya” is specified, and this is output to the speech synthesizer 17 as syllable information. Further, the roaring determination means 14 receives, for example, a key code corresponding to the “hi” kana character key 2a, a key code corresponding to the dakuten key 2b, and a “ya” yaw small letter key 2e from the character input device 2. When the corresponding key code is input at the same time, the stuttering character “Bya” is identified and output to the speech synthesizer 17 as syllable information. Further, the roaring determination means 14 may, for example, enter a key code corresponding to the “hi” kana character key 2a from the character input device 2, a key code corresponding to the semi-voice key 2c, and a “ya” yaw small letter key 2e. When the key code corresponding to is input at the same time, the stuttering character “Pya” is specified and output to the speech synthesizer 17 as syllable information.

  On the other hand, when the number of key codes input from the character input device 2 is one, or when two or more key codes are not input simultaneously from the character input device 2, the stuttering discrimination means 14 performs processing there. finish. Further, the roaring determination means 14 receives from the character input device 2 the kana character key 2a, a-line lower-case key 2d or ya-line of “ka line, sa line, ta line, na line, ha line, ma line, la line”. The process is also terminated when the key code corresponding to the lowercase key 2e is not input.

  As shown in FIG. 1, a key code is input from the character input device 2 to the stuttering determination unit 14. The stuttering discriminating unit 14 generates the syllable information of the stuttering character by the above-described method and outputs it to the voice synthesizing unit 17.

  The syllable character with prompting determination means 15 determines a syllable character with a prompting sound from the key code input from the character input device 2. That is, the syllable-with-prompting syllable discriminating means 15 monitors whether or not at least the key code of the kana character key 2a (see FIG. 2) and the key code of the tab line lowercase key 2f are input from the character input device 2, By detecting a state in which the key code of the key 2a and the lowercase key 2f are simultaneously pressed, the syllable character with prompting sound specified by these key codes is determined. Note that the syllable character with a prompt sound is a clear sound character with a prompt sound such as “hi”, a cloudy character character with a prompt sound such as “bi”, a semi-voice character character with a prompt sound such as “pi”, It means a syllable character including a stuttering character with a prompting sound such as “, Bya, Pyat”.

  Specifically, when the key code is input from the character input device 2, the syllable-with-prompting syllable discrimination means 15 determines whether the number of the key code is two or more, and the two or more key codes. It is determined whether or not are simultaneously input. Then, when the number of key codes is two or more and these are simultaneously input, the syllable-with-sound syllable discrimination means 15 refers to the correspondence table (not shown) and inputs two or more key codes. Is determined to correspond to at least the kana character key 2a and the tab row lowercase key 2f other than "O, N". Then, when the key code corresponds to at least the kana character key 2a and the tab row lowercase key 2f other than “O, N”, the syllable with phonic sound identifying means 15 identifies the syllable character with phonation indicated by these keys, Is output to the speech synthesizer 17 as syllable information.

  The syllable-with-sound syllable discrimination means 15 receives, for example, a key code corresponding to the “Hi” kana character key 2 a and a key code corresponding to the “T” lower-case key 2 f simultaneously from the character input device 2. In the case, the “hi” of the clear sound with prompting sound is specified, and this is output to the speech synthesizer 17 as syllable information. Further, the syllable-with-sound syllable discriminating means 15 includes, for example, a key code corresponding to the “Hi” kana character key 2 a from the character input device 2, a key code corresponding to the muddy key 2 b, When the key code corresponding to 2f is input at the same time, the voiced muffled character “bi” is identified and output to the speech synthesizer 17 as syllable information. Further, the syllable-with-sound syllable discriminating means 15 includes, for example, a key code corresponding to the “Hi” kana character key 2 a from the character input device 2, a key code corresponding to the semi-voiced sound key 2 c, and “ When the key code corresponding to the key 2f is input at the same time, the semi-voice character with prompt sound is identified and output to the speech synthesizer 17 as syllable information.

  In addition, the syllable-with-sound syllable discrimination means 15 includes, for example, a key code corresponding to the “hi” kana character key 2 a from the character input device 2, a key code corresponding to the “ya” YA row lowercase key 2 e, When the key code corresponding to the two-line lowercase key 2f is input at the same time, the “hya” of the stuttering stuttering character is specified and output to the speech synthesizer 17 as syllable information. The syllable-with-sound syllable discriminating means 15 is, for example, a key code corresponding to the “Hi” kana character key 2 a from the character input device 2, a key code corresponding to the muddy sound key 2 b, When the key code corresponding to 2e and the key code corresponding to the lower case key 2f of “” are input at the same time, the stuttering character “Byan” with the sound prompt is specified and this is used as the syllable information. The result is output to the combining means 17. Further, the syllable-with-sound syllable discriminating means 15 includes, for example, a key code corresponding to the “Hi” kana character key 2 a from the character input device 2, a key code corresponding to the semi-voice key 2 c, and a “ya” lowercase letter. When the key code corresponding to the key 2e and the key code corresponding to the lower case key 2f of “tsu” are input at the same time, the stuttering stuttering character “Pyat” is identified, and this is used as the syllable information. The result is output to the combining means 17.

  On the other hand, when the syllable-with-sound syllable discrimination means 15 has only one key code input from the character input device 2 or when two or more key codes have not been input simultaneously from the character input device 2, End the process. Further, the syllable-with-sound syllable discrimination means 15 also ends the process even when the key input corresponding to the kana character key 2a or the tab line lowercase key 2f other than “O,” is not input from the character input device 2. .

  As shown in FIG. 1, a key code is input from the character input device 2 to the syllable-with-prompt syllable discrimination means 15. Then, the syllable-with-prompting discrimination unit 15 generates syllable information of the syllable character with prompting sound by the above-described method, and outputs it to the speech synthesis unit 17.

  The voice waveform storage means 16 stores in advance the voice waveform of each syllable consisting of clear sound, sound repellent, semi-turbid sound, stuttering, and syllable with prompt sound. Specifically, the voice waveform storage means 16 is implemented by a memory, a hard disk or the like that can store data. As shown in FIG. 1, the speech waveform storage unit 16 outputs a speech waveform of each syllable including a clean sound, a sound repellent, a semi-turbid sound, a stuttering sound, and a syllable with a prompt sound to the speech synthesis unit 17.

  The voice synthesizing means 17 synthesizes voice according to the discrimination results of the above-mentioned discrimination means. That is, the speech synthesizing means 17 is discriminated from the speech waveform storing means 16 by each discriminating means comprising the clear sound repelling discrimination means 11, the muddy sound discrimination means 12, the semi-turbid sound discrimination means 13, the stuttering sound discrimination means 14 and the syllable discrimination syllable discrimination means 15. A voice is synthesized by extracting a speech waveform corresponding to each syllable character composed of a clear sound character, a repellent character, a semi-voice character, a stuttering character, and a syllable character with a prompt sound.

  For example, when the syllable information of the voice “ha” is input from the voice sound recognizing means 11, the voice synthesizer 17 extracts the voice waveform “ha” from the voice waveform storage means 16, and the voice output device 3 is output. Further, for example, when the syllable information of the muddy character “bi” is input from the muddy sound discriminating unit 12, the voice synthesizing unit 17 extracts the sound waveform of “bi” from the sound waveform storage unit 16 and outputs the sound. Output to device 3. For example, when the syllable information of the semi-voice character “pi” is input from the semi-turbid sound discrimination means 13, the speech synthesizer 17 extracts the speech waveform of “pi” from the speech waveform storage means 16. Output to the audio output device 3.

  For example, when the syllable information of the stuttering character “Fi” is input from the stuttering discrimination unit 14, the voice synthesis unit 17 extracts the voice waveform of “Fi” from the voice waveform storage unit 16. Output to the audio output device 3. For example, when the syllable information of “hi” is input from the syllable discriminating unit 15 with the prompt sound, the voice synthesizing unit 17 extracts the voice waveform of “hi” from the voice waveform storage unit 16. And output to the audio output device 3. As described above, the speech synthesizer 17 synthesizes speech at the timing when the key of the character input device 2 is pressed. Then, the voice output device 3 sequentially reproduces the voice based on the voice waveform input from the voice synthesizer 17.

  Syllable information is input to the speech synthesizer 17 from the clear sound repellent discrimination means 11, the muddy sound discrimination means 12, the semi-turbid sound discrimination means 13, the stuttering sound discrimination means 14, and the syllable discrimination syllable discrimination means 15. Then, the voice synthesizer 17 synthesizes the voice by the method described above and outputs it to the voice output device 3.

  As described above, the speech synthesizer 1 having the above-described configuration constantly monitors the key input status of the character input device 2 by each determination unit. Then, when any key of the character input device 2 is pressed by the operator and a key code corresponding to the key is input by the operator, the speech synthesizer 1 is syllable characters corresponding to the key code by each determining means. And the speech synthesizer 17 sequentially synthesizes speech corresponding to the determined syllable characters.

  Therefore, according to the speech synthesizer 1, since the speech corresponding to the input content of the character input device 2 is sequentially synthesized, the input completion button provided in the conventional speech synthesizer becomes unnecessary, and the speech is synthesized and reproduced. Can be shortened. Also, according to the speech synthesizer 1, only when the key code of the kana character key 2a and the key code of the dakuten key 2b or the like are simultaneously input from the character input device 2, the syllable such as muddy sound is synthesized. The speech synthesis is not erroneously synthesized, and the accuracy of speech synthesis can be improved.

[Speech synthesis output program]
Here, the speech synthesizer 1 can be realized by operating a general computer with a program that causes each of the above-described units and units to function. This program can be distributed via a communication line, or can be written on a recording medium such as a CD-ROM for distribution.

[Operation of Speech Synthesizer 1]
Hereinafter, the operation of the speech synthesizer 1 will be briefly described with reference to FIG. Here, the operation of the speech synthesizer 1 described below indicates the operation of the speech synthesizer 1 when the operator presses any key of the character input device 2 only once.

  First, the clear sound repellent discrimination means 11, the muddy sound discrimination means 12, the semi-turbid sound discrimination means 13, the stuttering sound discrimination means 14, and the syllable discrimination syllable discrimination means 15 are sent from the character input device 2 to the kana character key 2 a, the dakuten key 2 b, and the semi-dakuten key. It is monitored whether or not there is an input of the key codes of 2c, A-line lower-case key 2d, Y-line lower-case key 2e, and 2-line lower-case key 2f (step S1).

  Next, the key code corresponding to the key pressed by the operator is sent from the character input device 2 to the clear sound / acoustic sound discrimination means 11, the muddy sound discrimination means 12, the semi-turbid sound discrimination means 13, the stuttering sound discrimination means 14 and the syllable discrimination syllable discrimination means 15. When each is entered, each discriminating means corresponds to the syllable character of the input key code, which is a clear character, repellent character, muddy character, semi-voiced character, stuttering character or syllable character with prompt sound. Is determined (step S2).

  Next, the speech synthesizer 17 determines the syllable character determined by the clear sound repellent determination unit 11, the muddy sound determination unit 12, the semi-turbid sound determination unit 13, the stuttering determination unit 14, or the syllable determination unit 15 with prompt sound from the speech waveform storage unit 16. Is extracted and output to the audio output device 3 (step S3).

[Second Embodiment]
Hereinafter, the speech synthesizer 1A according to the second embodiment will be described in detail with reference to FIGS. The speech synthesizer 1A according to the second embodiment includes the repetitive signal output unit 18 and the repetitive waveform storage unit 19 shown in FIG. 4 and uses the character input device 2A instead of the character input device 2. Has the same configuration as the speech synthesizer 1 according to the first embodiment described above. Therefore, the same components as those in the speech synthesizer 1 described above are denoted by the same reference numerals and description thereof is omitted.

  The character input device 2A is characterized by outputting a make code and a break code to the speech synthesizer 1A in addition to the key code described above. Here, the makeup code is a code that is output when the operator presses a key of the character input device 2A, and indicates that the operator has pressed any key of the character input device 2A. The break code is a code that is output when the operator releases the key of the character input device 2A, and indicates that the operator has released any key of the character input device 2A. That is, during the period from when the make code is input from the character input device 2A to when the break code is input, one of the keys is continuously pressed by the operator (the state where the key is pressed). It means that.

  The repetitive signal output means 18 outputs a repetitive start signal or a repetitive end signal to the speech synthesizer 17 under a predetermined condition. Specifically, the repeat signal output means 18 monitors whether or not a make code and a break code have been input from the character input device 2A, so that the timing when the operator presses the key of the character input device 2A and the operation And the timing when the user releases the key of the character input device 2A.

  When the make code is input from the character input device 2 </ b> A, the repetition signal output unit 18 generates a repetition start signal and outputs it to the speech synthesis unit 17. Here, the repeat start signal is a voice to the voice synthesizer 17 when a makeup code is input from the character input device 2A (when an operator presses any key of the character input device 2A). It is a signal for instructing to repeatedly connect the repeated waveforms in the waveform repetition section.

  When the break code is input from the character input device 2 </ b> A, the repetitive signal output unit 18 generates a repetitive end signal and outputs it to the speech synthesis unit 17. Here, the repeat end signal means that the voice synthesizing unit 17 is voiced when a break code is input from the character input device 2A (when the operator releases any key of the character input device 2A). This is a signal for instructing to end the connection of the repeated waveform to the waveform.

  As shown in FIG. 4, the repeat signal output means 18 receives a make code and a break code from the character input device 2A. Then, the repetitive signal output unit 18 generates a repetitive start signal and a repetitive end signal by the above-described method, and outputs them to the speech synthesis unit 17.

  The repetitive waveform storage means 19 stores in advance a repetitive waveform indicating a waveform for several cycles in a section corresponding to a vowel among speech waveforms of syllables including clear sound, repellent sound, muddy sound, semi-turbid sound, stuttering, and syllable with prompt sound. Is. Hereinafter, the repeated waveform will be briefly described with reference to FIG. Note that the audio waveform of FIG. 5A is exaggerated for the sake of convenience. Further, the voice waveform shown in FIG. 5A is merely an example, and the amplitude, period, and length of the waveform are not limited to those shown in FIG.

  For example, the voice waveform (100 msec) of the clear sound “ka” shown in FIG. 5A includes a waveform corresponding to the consonant “k” (30 msec) and a waveform corresponding to the vowel “a” ( 70 msec). As shown in FIG. 5A, the waveform (70 msec) of the portion corresponding to “a” of this vowel is composed of a plurality of waveforms (5 msec) having similar amplitudes and fundamental periods. A waveform corresponding to a period is called a repetitive waveform here.

  Specifically, the repetitive waveform storage means 19 is implemented by a memory, a hard disk or the like that can store data. As shown in FIG. 4, the repetitive waveform storage unit 19 outputs a repetitive waveform of each syllable composed of a clear sound, a sound repellent, a semi-turbid sound, a stuttering sound, and a syllable with a prompt sound, to the speech synthesis unit 17.

  In the speech synthesizer 1A including the repetitive signal output means 18 and the repetitive waveform storage means 19, the speech synthesizer 17 extracts a speech waveform from the speech waveform storage means 16 in accordance with the syllable information input from each discrimination means. And, in accordance with the repetition start signal from the repetition signal output means 18, the repetition waveform corresponding to the speech waveform described above is extracted from the repetition waveform storage means 19. Then, the speech synthesizer 17 synthesizes the speech by repeatedly connecting the repetitive waveforms in the repetitive sections of the speech waveform until the repetitive end signal is input from the repetitive signal output unit 18.

  For example, as shown in FIG. 5B, the voice synthesizing unit 17 reads the voice character from the voice waveform storage unit 16 when the operator continuously presses the “ka” key of the character input device 2A for 1000 msec. The voice waveform of “ka” is extracted, and the repeated waveform of the “ka” of the clear character is extracted from the repeated waveform storage means 19. Then, the voice synthesizer 17 synthesizes the voice by connecting the “m” repeated waveform of 5 msec 186 times after the “m” voice waveform of 70 msec. The synthesized voice of “ka” having undergone such processing is reproduced as “ka” when the vowel “a” is lengthened and output by the voice output device 3. When the operator releases the “ka” key of the character input device 2A at 1000 msec, the speech synthesizer 17 generates a waveform corresponding to the remaining 30 msec of the repetitive section as shown in FIG. Further connection is made after the repetitive waveform described above. However, in this case, the amplitude of the remaining waveform in the repetitive section is rapidly reduced to the level of −20 db, and is terminated at 50 msec.

  In the speech synthesizer 1A having the above-described configuration, the repeat signal output unit 18 constantly monitors the input of the make code and the break code from the character input device 2A, so that the operator can select one of the character input devices 2A. It is recognized whether the key is pressed or released. The voice synthesizer 1A then waits for the break code to be input after any key of the character input device 2A is pressed by the operator (after the make code is input) until the key is released. Until the operator presses the key of the character input device 2A, the voice synthesizing unit 17 synthesizes the voice having a longer sound length by repeatedly connecting the repeated waveforms in the repeated section of the voice waveform.

  Therefore, according to the speech synthesizer 1A, when the operator continues to press any key of the character input device 2A, the syllable corresponding to the key can be synthesized with a longer syllable. The variation of can be increased.

[Operation of Speech Synthesizer 1A]
Hereinafter, the operation of the speech synthesizer 1A will be briefly described with reference to FIG. Here, the operation of the speech synthesizer 1A described below indicates the operation of the speech synthesizer 1A when the operator presses any key of the character input device 2A only once.

  First, the clear sound repelling sound discrimination means 11, the muddy sound discrimination means 12, the semi-turbid sound discrimination means 13, the stuttering sound discrimination means 14, and the syllable discrimination syllable discrimination means 15 are sent from the character input device 2A to the kana character key 2a, the muddy point key 2b, and the semi-sound mark key. It is monitored whether or not there is an input of the key codes of 2c, A-line lower-case key 2d, Y-line lower-case key 2e, and 2-line lower-case key 2f (step S11). Further, the repetition signal output means 18 monitors whether or not a make code and a break code are input from the character input device 2A (step S12).

  Next, from the character input device 2A, the keys pressed by the operator to the clear sound repellent determination means 11, the muddy sound determination means 12, the semi-turbid sound determination means 13, the stuttering sound determination means 14, and the syllable determination syllable determination means 15 respectively. When the corresponding key code is input, each discriminating means corresponds to the syllable character of the clear key character, the sound-repellent character, the muddy character, the semi-voiced character, the stuttering character, or the syllable character with prompting sound. (Step S13).

  When a make code indicating that the operator has pressed a key is input from the character input device 2A to the repeat signal output means 18 (Yes in step S14), the repeat signal output means 18 repeats. A start signal is generated and output to the speech synthesizer 17 (step S15). On the other hand, if the above-described make code is not input from the character input device 2A to the repetitive signal output means 18, the process returns to step S12 to continue monitoring.

  Next, the speech synthesizer 17 determines the syllable character determined by the clear sound repellent determination unit 11, the muddy sound determination unit 12, the semi-turbid sound determination unit 13, the stuttering determination unit 14, or the syllable determination unit 15 with prompt sound from the speech waveform storage unit 16. A speech waveform corresponding to is extracted (step S16). Further, the speech synthesizer 17 applies the syllable character determined by the repetitive waveform storage unit 19 by the clear sound repellent determination unit 11, the muddy sound determination unit 12, the semi-turbid sound determination unit 13, the stuttering determination unit 14, or the syllable determination unit 15 with prompt sound A corresponding repeated waveform is extracted (step S17).

  Next, the speech synthesizer 17 repeatedly connects the repetitive waveforms in the repetitive section of the speech waveform (step S18). Next, when a break code indicating that the operator has released the key is input from the character input device 2A to the repeat signal output means 18 (Yes in step S19), the repeat signal output means 18 A repeat end signal is generated and output to the speech synthesizer 17 (step S20). On the other hand, if the break code is not input from the character input device 2A to the repetitive signal output means 18, the process returns to step S18 to continue monitoring (No in step S19). Next, the speech synthesizer 17 ends the connection of the repeated waveforms, and outputs the synthesized speech to the speech output device 3 (step S21).

[Third Embodiment]
Hereinafter, the speech synthesizer 1B according to the third embodiment will be described in detail with reference to FIG. The speech synthesizer 1B according to the third embodiment is the same as the speech synthesizer 1A according to the second embodiment described above except that it includes the repetitive waveform generating means 20 shown in FIG. It has the composition of. Therefore, about the structure which overlaps with above-described speech synthesis apparatus 1A, the same code | symbol is attached | subjected and description is abbreviate | omitted. The description of the operation of the speech synthesizer 1B is also omitted.

  The repetitive waveform generating means 20 generates a repetitive waveform from the speech waveform stored in the speech waveform storage means 16. First, the repetitive waveform generating means 20 applies a pitch mark to a speech waveform stored using the technique disclosed in Japanese Patent Application Laid-Open No. 2003-122380 “Pitch Mark Appending Device, Processing Method Therefor, and Storage Medium”. Divide into Next, the cross-correlation with the preceding and following fundamental cycle units is calculated for all the divided fundamental cycle units. Finally, a basic period unit having the highest sum of the cross correlations before and after is extracted, and that portion is determined as a repetitive waveform.

  For example, the repetitive waveform generation means 20 can generate 14 pitch marks when 15 pitch marks can be added to the waveform (70 msec) corresponding to “a” of the vowel “a” in FIG. The correlation before and after is calculated for all the basic period units, and the portion with the highest correlation is extracted as a repeated waveform. In this way, by using a portion having the highest correlation, that is, a stable waveform as a repetitive waveform, the audio output device 3 can output stably.

  As shown in FIG. 7, the repetitive waveform generation means 20 receives a repetitive waveform generation instruction, which is an instruction to generate a repetitive waveform, from the speech synthesis means 17, and a speech waveform from the speech waveform storage means 16. Entered. Then, the repetitive waveform generation means 20 generates a repetitive waveform by the method described above and outputs it to the speech synthesis means 17.

  In the speech synthesizer 1B including the repetitive waveform generating unit 20 as described above, the speech synthesizing unit 17 extracts a speech waveform from the speech waveform storing unit 16 according to the syllable information from each determining unit, and the repetitive signal output unit 18. Is output to the repetitive waveform generation means 20 in accordance with the repetitive start signal from the above. The speech synthesizer 17 receives the repetitive waveform from the repetitive waveform generator 20 and repeatedly connects the repetitive waveforms in the repetitive section of the speech waveform until the repetitive end signal is input from the repetitive signal output unit 18. Synthesize speech.

  In the speech synthesizer 1B having the above-described configuration, the repeat signal output unit 18 constantly monitors the input of the make code and the break code from the character input device 2A, so that the operator can select one of the character input devices 2A. It is recognized whether the key is pressed or released. The voice synthesizer 1B then waits for the break code to be input after any key of the character input device 2A is pressed by the operator (after the makeup code is input) until the key is released. Until the operator presses the key of the character input device 2A, the voice synthesizing unit 17 synthesizes the voice having a longer sound length by repeatedly connecting the repeated waveforms in the repeated section of the voice waveform.

  Therefore, according to the speech synthesizer 1B, when the operator continues to press any key of the character input device 2A, the syllable corresponding to the key can be synthesized with a longer syllable. The variation of can be increased.

[Fourth Embodiment]
The speech synthesizer 1C according to the fourth embodiment will be described in detail below with reference to FIGS. The speech synthesizer 1C according to the fourth embodiment includes the accent signal output means 21 shown in FIG. 8, and instead of the speech waveform storage means 16, the low speech waveform storage means 22 shown in FIG. And a configuration similar to that of the speech synthesizer 1A according to the second embodiment described above except that the character input device 2B is used instead of the character input device 2A. Therefore, about the structure which overlaps with above-described speech synthesis apparatus 1A, the same code | symbol is attached | subjected and description is abbreviate | omitted.

  As shown in FIG. 9, the character input device 2B has an accent descending function in addition to a kana character key 2a, a dakuten key 2b, a semi-dakuten key 2c, a line lowercase key 2d, a line lowercase key 2e, and a line lowercase key 2f. A key 2g, an accent raising key 2h, and an accent delimiter key 2i are further provided.

  The accent signal output means 21 outputs an accent fall signal, an accent rise signal, or an accent delimiter signal to the speech synthesizer 17 under predetermined conditions. Specifically, the accent signal output means 21 is supplied from the character input device 2B, from the kana character key 2a, the dakuten key 2b, the semi-dakuten key 2c, the a lower case key 2d, the yaw lower case key 2e, and the ta row lower case key 2f. It is monitored whether the key code of the accent lowering key 2g, the accent rising key 2h, or the accent delimiter key 2i is input before the key code of each character key.

  Then, when the key code of the accent lowering key 2g is input from the character input device 2B, the accent signal output means 21 generates an accent lowering signal and outputs it to the speech synthesizer 17. Here, the accent lowering signal is a signal for lowering the voice accent synthesized by the voice synthesizing means 17. That is, when an accent lowering signal is input to the speech synthesizing means 17, the speech synthesizing means 17 receives a syllable character determined by each determining means after the accent decreasing signal is input from the low-speech waveform storage means 22 described later. A low speech waveform corresponding to is extracted.

  For example, when the operator inputs the characters “sa”, “ku”, and “ra” with the character input device 2B, if the operator presses the accent lowering key 2g before pressing the “sa” kana character key 2a, The speech synthesizer 17 extracts the low speech waveform corresponding to the syllable characters “sa”, “ku”, and “ra” from the low speech waveform storage 22. That is, low accent sounds such as “sa (low)”, “ku (low)”, and “ra (low)” are synthesized.

  The accent signal output means 21 generates an accent increase signal when the key code of the accent increase key 2h is input from the character input device 2B, and outputs this to the speech synthesizer 17. Here, the accent increase signal is a signal for increasing the accent of the voice synthesized by the voice synthesizing means 17. That is, when an accent increase signal is input to the speech synthesizer 17, the speech synthesizer 17 receives the syllable character determined by each determination unit after the accent increase signal is input from the high speech waveform storage unit 23 described later. Extract high speech waveform corresponding to.

  For example, when the operator inputs the characters “sa”, “ku”, and “ra” with the character input device 2B, if the accent up key 2h is pressed before the character key 2a of “sa” is pressed, The speech synthesizer 17 extracts a high speech waveform corresponding to each syllable character of “sa”, “ku”, and “ra” from the high speech waveform storage unit 23. That is, high accent sounds such as “sa (high)”, “ku (high)”, and “ra (high)” are synthesized.

  The accent signal output means 21 generates an accent delimiter signal when the key code of the accent delimiter key 2i is input from the character input device 2B, and outputs this to the speech synthesizer 17. Here, the accent delimiter signal is a signal for delimiting the accent of the speech synthesized by the speech synthesizer 17 for each accent phrase, for example. That is, when an accent descending signal is input to the speech synthesizer 17, the speech synthesizer 17 receives the first syllable character determined by each determining unit after the accent delimiter signal is input from the low speech storage unit 22. A low speech waveform corresponding to is extracted. Then, a high speech waveform corresponding to the second and subsequent syllable characters discriminated by each discriminating unit after the accent delimiter signal is input is extracted from the high speech storage unit 23.

  For example, when the operator inputs the characters “sa”, “ku”, and “ra” using the character input device 2B, if the accent delimiter key 2i is pressed before pressing the “s” kana character key 2a, The voice synthesizing unit 17 extracts a low voice waveform corresponding to the syllable character “sa” from the low voice waveform storage unit 22, and outputs a high voice waveform corresponding to each syllable character “ku” and “ra” to the high voice. Extracted from the waveform storage means 23. That is, voices in which low accent and high accent are mixed, such as “sa (low)”, “ku (high)”, and “ra (high)” are synthesized.

  The low speech waveform storage means 22 stores a low speech waveform indicating a low accent speech waveform of each syllable composed of clear sound, repellent sound, muddy sound, semi-turbid sound, stuttering, and syllable with prompt sound. Specifically, the low audio waveform storage means 22 is implemented by a memory, a hard disk, or the like that can store data. As shown in FIG. 8, the low speech waveform storage unit 22 outputs low speech waveforms of clean sound, sound repellent, semi-turbid sound, stuttering, and syllable with phonation to the speech synthesis unit 17.

  The high speech waveform storage means 23 stores a high speech waveform indicating a high accent speech waveform of each syllable consisting of clear sound, repellent sound, muddy sound, semi-turbid sound, stuttering, and syllable with prompt sound. Specifically, the high speech waveform storage means 23 is implemented by a memory, a hard disk or the like that can store data. As shown in FIG. 8, the high speech waveform storage unit 23 outputs high speech waveforms of clean sound, repellent sound, semi-turbid sound, stuttering sound, and syllable with prompt sound to the speech synthesis unit 17.

  In the speech synthesizer 1C including the accent signal output means 21, the low speech waveform storage means 22 and the high speech waveform storage means 23, the speech synthesizer 17 includes the syllable information from each determination means and the accent signal output means 21. A low speech waveform or a high speech waveform is extracted from the low speech waveform storage means 22 or the high speech waveform storage means 23 in accordance with the accent fall signal, accent rise signal or accent delimiter signal. Further, the speech synthesizer 17 extracts a repetitive waveform corresponding to the low speech waveform or the high speech waveform from the repetitive waveform storage unit 19 in accordance with the repetitive start signal from the repetitive signal output unit 18. Then, the speech synthesizer 17 synthesizes the speech by repeatedly connecting the repetitive waveforms in the repetitive sections of the low speech waveform or the high speech waveform until the repetitive end signal is input from the repetitive signal output unit 18.

  Note that in the speech synthesizer 1C, if any signal of the accent lowering signal, the accent rising signal or the accent delimiter signal is not input from the accent signal output unit 21 to the speech synthesizing unit 17, the speech synthesizing unit 17 For the first character, the low speech waveform corresponding to the syllable character determined by each determination means is extracted from the low speech waveform storage means 22, and after the second character, each determination means is determined from the high speech waveform storage means 23. A high speech waveform corresponding to the syllable character is extracted. As described above, in the speech synthesizer 1C, the speech synthesizer 17 counts the number of syllable characters determined by each determining unit by a counting unit (not shown). Moreover, in the speech synthesizer 1C, as described above, the repetitive waveform storage unit 19 stores in advance repetitive waveforms of both the low speech waveform and the high speech waveform.

  The speech synthesizer 1C having the above configuration constantly monitors the input status of the accent lowering key 2g, accent increasing key 2h, or accent delimiter key 2i of the character input device 2B by the accent signal output means 21. When the operator presses the accent lowering key 2g, accent rising key 2h, or accent delimiter key 2i of the character input device 2B and the key code corresponding to the key is input, the voice synthesizing unit 1C 17 extracts a low speech waveform or a high speech waveform according to the key code, and sequentially synthesizes speech with accents.

  Therefore, according to the speech synthesizer 1C, when the operator presses the accent lowering key 2g, accent rising key 2h or accent delimiter key 2i of the character input device 2B, the syllable accent can be changed. Variations can be further increased.

[Operation of Speech Synthesizer 1C]
Hereinafter, the operation of the speech synthesizer 1C will be briefly described with reference to FIG. Here, the operation of the speech synthesizer 1C described below indicates the operation of the speech synthesizer 1C when the operator presses any key of the character input device 2B only once.

  First, the accent signal output means 21 monitors whether or not the key code of the accent lowering key 2g, the accent raising key 2h or the accent delimiter key 2i is input from the character input device 2B, and the clear sound and sound repelling judgment means 11, The muddy sound discriminating means 12, the semi-turbid sound discriminating means 13, the stuttering sound discriminating means 14 and the syllable discriminating means 15 with the prompt sound are sent from the character input device 2B to the kana character key 2a, the dakuten key 2b, the semi-dakuten key 2c, It is monitored whether or not there is an input of the key codes of the YA row lower case key 2e and the TA row lower case key 2f (step S21). Further, the repeat signal output means 18 monitors whether or not a make code and a break code are input from the character input device 2B (step S22).

  Next, when the key code of the accent lowering key 2g, the accent increasing key 2h or the accent delimiter key 2i pressed by the operator is input from the character input device 2B to the accent signal output means 21, the accent signal output means 21 generates an accent drop signal, an accent rise signal, or an accent delimiter signal and outputs it to the speech synthesizer 17 (step S23).

  Next, a key code corresponding to the key pressed by the operator is input to each of the clear sound repellent determination means 11, the muddy sound determination means 12, the semi-turbid sound determination means 13, the stuttering sound determination means 14, and the syllable determination syllable determination means 15. Then, each discrimination means discriminates whether the input key code corresponds to a syllable character of a clear sound character, a sound-repellent character, a cloudy character, a semi-voiced character, a stuttering character, or a syllable character with a prompt sound. (Step S24).

  When a make code indicating that the operator has pressed a key is input from the character input device 2B to the repeat signal output means 18 (Yes in step S25), the repeat signal output means 18 repeats. A start signal is generated and output to the speech synthesizer 17 (step S26). On the other hand, if the above-described make code is not input from the character input device 2B to the repetitive signal output means 18, the process returns to step S22 and monitoring is continued.

  Next, the speech synthesizer 17 receives the clear sound repellent discrimination means 11, the muddy sound discrimination means 12, the semi-turbid sound discrimination means 13, the stuttering sound discrimination means 14, or the syllable discrimination with prompt sound from the low voice waveform storage means 22 or the high voice waveform storage means 23. A low speech waveform or a high speech waveform corresponding to the syllable character discriminated by each discriminating means comprising means 15 is extracted (step S27). Further, the speech synthesizer 17 is discriminated from the repetitive waveform storage unit 19 by each discriminating unit consisting of the clear sound repellent discriminating unit 11, the muddy sound discriminating unit 12, the semi-turbid sound discriminating unit 13, the stuttering sound discriminating unit 14, or the syllable discriminating unit 15 with prompt sound. A repeated waveform of a low speech waveform or a high speech waveform corresponding to the syllable character is extracted (step S28).

  Next, the speech synthesizer 17 repeatedly connects the repetitive waveforms in the repetitive section of the low speech waveform or the high speech waveform (step S29). Next, when a break code indicating that the operator has released the key is input from the character input device 2B to the repetitive signal output means 18 (Yes in step S30), the repetitive signal output means 18 A repeat end signal is generated and output to the speech synthesizer 17 (step S31). On the other hand, if the break code is not input from the character input device 2B to the repetitive signal output means 18, the process returns to step S29 and continues monitoring (No in step S30). Next, the speech synthesizer 17 ends the repetitive waveform connection and outputs the synthesized speech to the speech output device 3 (step S32).

[Specific example of speech synthesis using speech synthesizer 1C]
Hereinafter, a specific example of speech synthesis using the speech synthesizer 1C will be briefly described. In the following example, a case where the speech synthesizer 1C is used to synthesize an accent voice of “I (low), (high), (high), (low)” will be described.

  In this case, the operator first presses the “I” kana character key 2a (see FIG. 9) of the character input device 2B. Then, the character input device 2 </ b> B outputs the key code of the “I” kana character key 2 a to each determination unit and also outputs the make code to the repetitive signal output unit 18. Next, the clean sound repellent determination means 11 determines that the key code is “i” of the clear sound character, and outputs this to the speech synthesis means 17 as syllable information. The repeat signal output means 18 generates a repeat start signal according to the make code and outputs it to the speech synthesizer 17.

  Next, the voice synthesizing unit 17 extracts the low voice waveform of “I” from the low voice waveform storage unit 22 according to the syllable information, and the low level of “I” from the repeated waveform storage unit 19 according to the repetition start signal. Extract repeated waveforms of speech waveform. Then, the speech synthesizer 17 connects the repetitive waveforms described above in the repetitive section of the low speech waveform “I” until the repetitive end signal is input from the repetitive signal output unit 18.

  Next, the operator simultaneously presses the “t” row lowercase key 2f and the “ra” kana character key 2a of the character input device 2B. Then, the character input device 2B outputs the key code of the “t” row lowercase key 2f and the key code of the “ra” kana character key 2a to each discriminating means, and the repeated signal output means. For 18, the make code is output. Next, the syllable-with-sound syllable determination unit 15 determines that the key code is “Raku” as a clear sound character with a prompt sound, and outputs this to the speech synthesis unit 17 as syllable information. The repeat signal output means 18 generates a repeat start signal according to the make code and outputs it to the speech synthesizer 17.

  Next, the speech synthesizer 17 extracts the high speech waveform of “Raku” from the high speech waveform storage device 23 according to the syllable information, and “Rak” from the repeated waveform storage device 19 according to the repetition start signal. Extract repeated waveform of high voice waveform. Then, the speech synthesizer 17 connects the repetitive waveforms described above in the repetitive section of “Raku” high speech waveform until the repetitive end signal is input from the repetitive signal output unit 18.

  Next, the operator simultaneously presses the YA row lower case key 2e and the “K” character key 2a of “YA” of the character input device 2B. Then, the character input device 2B outputs the key code of the character key 2e and the key code of the character key 2a of "Shi" to each discrimination means, and the repeated signal output means. For 18, the make code is output. Next, the stuttering discrimination means 14 discriminates that the key code is a stuttering character “sha” and outputs it to the speech synthesis means 17 as syllable information. The repeat signal output means 18 generates a repeat start signal according to the make code and outputs it to the speech synthesizer 17.

  Next, the voice synthesizing means 17 extracts the high voice waveform of “sha” from the high voice waveform storage means 23 according to the syllable information, and the high value of “sha” from the repeated waveform storage means 19 according to the repetition start signal. Extract repeated waveforms of speech waveform. Then, the speech synthesizer 17 connects the repetitive waveforms described above in the repetitive section of the high speech waveform of “sha” until the repetitive end signal is input from the repetitive signal output unit 18.

  Next, the operator presses the accent lowering key 2g of the character input device 2B. Then, the accent signal output means 21 outputs an accent lowering signal to the speech synthesis means 17.

  Next, the operator presses the character key 2a of “I” on the character input device 2B. Then, the character input device 2 </ b> B outputs the key code of the “I” kana character key 2 a to each determination unit and also outputs the make code to the repetitive signal output unit 18. Next, the clean sound repellent determination means 11 determines that the key code is “i” of the clear sound character, and outputs this to the speech synthesis means 17 as syllable information. The repeat signal output means 18 generates a repeat start signal according to the make code and outputs it to the speech synthesizer 17.

  Next, the speech synthesizer 17 extracts the low voice waveform of “I” from the low voice waveform storage unit 22 according to the syllable information and the accent lowering signal, and the repeated waveform storage unit 19 according to the repetition start signal. To extract a repeated waveform of the low voice waveform of “I”. Then, the speech synthesizer 17 connects the repetitive waveforms described above in the repetitive section of the low speech waveform “I” until the repetitive end signal is input from the repetitive signal output unit 18.

  As described above, in the speech synthesizer 1C, the operator uses the keys of the character input device 2B (see FIG. 9) as “I”, “tsu” + “ra”, “nya” + “shi”, and the accent lowering key. By pressing in the order of 2 g and “I”, it is possible to appropriately synthesize an accented voice of “I (low), (high), (high), (low)”. Note that when the speech synthesizer 1C synthesizes speech with an accent such as “I (high), (low), (low), low (low)”, the operator uses the keys of the character input device 2B (FIG. 9). Reference) may be pressed in the order of the accent up key 2h, “i”, accent down key 2g, “tsu” + “ra”, “nya” + “shi”, “i”.

[Fifth Embodiment]
Hereinafter, a speech synthesizer 1D according to the fifth embodiment will be described in detail with reference to FIG. The speech synthesizer 1D according to the fifth embodiment is the same as the speech synthesizer 1C according to the fourth embodiment described above except that it includes the repetitive waveform generation means 20 shown in FIG. It has the composition of. Therefore, about the structure which overlaps with above-described speech synthesis apparatus 1C, the same code | symbol is attached | subjected and description is abbreviate | omitted. The description of the operation of the speech synthesizer 1D is also omitted.

  The repetitive waveform generating unit 20 repeats a low speech waveform or a high speech waveform from a low speech waveform stored in the low speech waveform storage unit 22 or a high speech waveform stored in the high speech waveform storage unit 23. A waveform is generated. For example, the waveform (70 msec) corresponding to “a” of the vowel “ka” in FIG. 5A is compared with the waveforms before and after the basic period unit in which the pitch mark can be added, and the correlation is obtained. The highest part is extracted as a repeated waveform. In this way, by using a portion having the highest correlation, that is, a stable waveform as a repetitive waveform, the audio output device 3 can output stably.

  As shown in FIG. 11, the repetitive waveform generation means 20 receives a repetitive waveform generation instruction, which is an instruction to generate a repetitive waveform, from the speech synthesizer 17, and the low sound waveform storage means 22 or high sound. A low speech waveform or a high speech waveform is input from the waveform storage means 23. Then, the repetitive waveform generation means 20 generates a low speech waveform or a high speech waveform repetitive waveform by the method described above, and outputs this to the speech synthesis means 17.

  In the speech synthesizer 1D provided with such a repetitive waveform generating means 20, the speech synthesizer 17 receives the low speech waveform or the low speech waveform from the low speech waveform storage means 22 or the high speech waveform storage means 23 according to the syllable information from each discrimination means. An instruction to extract a high speech waveform and to instruct the repetitive waveform generation unit 20 to indicate the low speech waveform or the repetitive waveform corresponding to the high speech waveform in accordance with the repetitive start signal from the repetitive signal output unit 18 A repetitive waveform generation instruction is output. The speech synthesizer 17 receives the low speech waveform or the high speech waveform repetitive waveform from the repetitive waveform generator 20, and the low speech waveform or high speech waveform until the repetitive end signal is input from the repetitive signal output unit 18. A speech is synthesized by repeatedly connecting a low speech waveform or a high speech waveform in a repetitive section.

  The speech synthesizer 1D having the above configuration constantly monitors the input of the make code and break code from the character input device 2B by the repetitive signal output means 18, so that the operator can select either of the character input devices 2B. It is recognized whether the key is pressed or released. At the same time, the speech synthesizer 1D constantly monitors the input status of the accent lowering key 2g, accent increasing key 2h or accent delimiter key 2i of the character input device 2B by the accent signal output means 21.

  When the operator presses the accent lowering key 2g, the accent rising key 2h, or the accent delimiter key 2i of the character input device 2B and the key code corresponding to the key is input, the voice synthesizing unit 1D 17 extracts a low voice waveform or a high voice waveform according to the key code. At the same time, the speech synthesizer 1D is configured so that any one of the keys of the character input device 2B is pressed by the operator (after the make code is input) until any one of the keys is released (the break code is input). Until it is done, the speech synthesis means 17 repeatedly connects the low speech waveform or the high speech waveform repetitive waveform in the repetitive section of the low speech waveform or the high speech waveform. In this way, the speech synthesizer 1D synthesizes the accented speech sequentially while making the sound longer only while the operator is pressing the key of the character input device 2B.

  Therefore, according to the speech synthesizer 1D, when the operator presses the accent lowering key 2g, the accent increasing key 2h or the accent delimiter key 2i of the character input device 2B, the syllable accent can be changed and the operator can When one of the keys of the character input device 2B is continuously pressed, the syllable corresponding to the key can be synthesized with a longer sound, so that variations in speech synthesis can be increased.

  The speech synthesizer and the speech synthesizer program according to the present invention have been specifically described above by the embodiments for carrying out the invention. However, the gist of the present invention is not limited to these descriptions, and the scope of the claims Should be interpreted broadly based on the description. Needless to say, various changes and modifications based on these descriptions are also included in the spirit of the present invention.

  For example, in the above-described speech synthesizer 1, 1A, 1B, 1C, 1D, the operator uses the kana character key 2a of the character input devices 2, 2A, 2B, the muddy sound key 2b, the semi-sound sound key 2c, and the lower-case key. 2d, YA row lowercase key 2e, or TA row lowercase key 2f are combined to synthesize a voice, but before kana character key 2a, muddy sound key 2b, semi-turbid sound key 2c, a line A configuration may be adopted in which speech is synthesized when the lowercase key 2d, the lowercase key 2e, or the lowercase key 2f is pressed.

  Further, in the speech synthesizer 1, 1A, 1B, 1C, 1D, the operator uses the kana character key 2a, the muddy sound key 2b, the semi-sound key 2c, and the lowercase letter key 2d of the character input devices 2, 2A, 2B. , The accented voice is synthesized when the accent lowering key 2g, the accent raising key 2h, or the accent delimiter key 2i is pressed before the YA row lowercase key 2e or the TA row lowercase key 2f. The character key 2a, the muddy sound key 2b, the semi-voiced sound key 2c, the a-line lower-case key 2d, the ya-line lower-case key 2e or the t-line lower-case key 2f, the accent-decreasing key 2g, the accent-rising key 2h, or the accent-separating key 2i It is good also as a structure which synthesize | combines the audio | voice which added the accent, when pressed.

1, 1A, 1B, 1C, 1D Speech synthesizer 2, 2A, 2B Character input device 2a Kana character key 2b Cloud point key 2c Half cloud point key 2d A line lower case key 2e Ya line lower case key 2f Ta line lower case key 2g 2h Accent rise key 2i Accent break key 3 Voice output device 11 Clear sound repellent discrimination means 12 Muddy sound discrimination means 13 Semi-turbid sound discrimination means 14 Stuttering sound discrimination means 15 Stimulus syllable discrimination means 16 Speech waveform storage means 17 Speech synthesis means 18 Repeat signal output means 19 Repetitive waveform storage means 20 Repetitive waveform generation means 21 Accent signal output means 22 Low speech waveform storage means 23 High speech waveform storage means

Claims (2)

This is a speech synthesizer that synthesizes speech corresponding to the input contents of a character input device having character keys consisting of a kana character key, a cloud point key, a semi-cloud point key, a line lowercase key, a line lowercase key, and a line lowercase key. And
By detecting a state in which the kana character key is pressed, a sound and sound recognizing means for determining a sound character or a sound repellent character specified by a key code of the kana character key;
A muddy sound discriminating means for determining a muddy character identified by a key code of the kana character key and the dakuten key by detecting a state in which the kana character key and the dakuten key are simultaneously pressed;
A semi-turbid sound discriminating means for discriminating a semi-turbid character identified by a key code of the kana character key and the semi-turbid point key by detecting a state in which the kana character key and the semi-turbid point key are simultaneously pressed;
By detecting the state in which the kana character key and the lowercase letter key are simultaneously pressed, or the state in which the kana character key and the lowercase key are simultaneously pressed, the kana character key and the arrow key are detected. A stuttering discriminating means for discriminating a stuttering character specified by a key code of a row lowercase key or the key code of the kana character key and the yaw lowercase key;
A syllable with a timbre that discriminates a syllable character with a timbre identified by the key code of the kana character key and the lowercase letter key by detecting a state in which the kana letter key and the lowercase letter key are simultaneously pressed. Discrimination means;
Kiyoshi sound, repellent sound, turbid sound, semi-dull sound, sound waveform storage means for storing voice waveform of each syllable consisting screw sound and prompting audio-syllable,
From the speech waveform storage means, the clear sound repellent discrimination means, the muddy sound discrimination means, the semi-turbid sound discrimination means, the stuttering sound discrimination means, and the singing syllable discrimination means, the clear sound characters and the sound repellent includes characters, the voiced character, the semi-voiced character, and a speech synthesis means for synthesizing speech by extracting the voice waveform corresponding to each syllable character consisting of the contracted sound character and the double consonant with syllabary,
In addition to the key code, a make code indicating that each key has been pressed and a break code indicating that each key has been released are input from the character input device.
Whether or not the make code and the break code are input from the character input device is monitored, and when the make code is input, a repeat start signal is generated and output to the speech synthesizer, When a break code is input, a repetition signal output unit that generates a repetition end signal and outputs the signal to the speech synthesis unit;
Repetitive waveform storage means for storing a repetitive waveform indicating a waveform of a certain period in a section corresponding to a vowel among speech waveforms of each syllable,
The speech synthesis means extracts the speech waveform corresponding to each syllable character from the speech waveform storage means, and when the repetition start signal is input from the repetition signal output means, the repetition waveform storage The repetitive waveform corresponding to each syllable character is extracted from the means, and the repetitive waveform is repeatedly connected in the repetitive section of the speech waveform until the repetitive end signal is input from the repetitive signal output means. As well as
In the case where the character input device further includes an accent lowering key, an accent increasing key, and an accent delimiter key,
Whether or not the key code of the accent lowering key, the key code of the accent rising key or the key code of the accent delimiter key is input from the character input device before the key code of each character key; When the key code of the accent descending key is input, an accent descending signal is generated and output to the speech synthesizer, and when the key code of the accent increasing key is input, an accent increasing signal is generated. Output to the speech synthesizer, and when the key code of the accent separator key is input, the system includes an accent signal output unit that generates an accent separator signal and outputs it to the speech synthesizer.
The voice waveform storage means includes
Low speech waveform storage means for storing a low speech waveform indicating a low accent speech waveform of each syllable;
A high voice waveform storage means for storing a high voice waveform indicating a high accent voice waveform of each syllable;
The speech synthesis means
When the accent lowering signal is input from the accent signal output means, the syllable characters corresponding to the syllable characters determined by the determining means after the accent lowering signal is input from the low speech waveform storage means. Synthesize speech by extracting low speech waveform,
When the accent increase signal is input from the accent signal output means, the syllable characters corresponding to the syllable characters determined by the determination means after the accent increase signal is input from the high speech waveform storage means. Synthesize speech by extracting high speech waveform,
When the accent separation signal is input from the accent signal output means, the first syllable character determined by the determination means after the accent separation signal is input from the low speech waveform storage means The corresponding low speech waveform is extracted, and the high speech corresponding to the second and subsequent syllable characters determined by the determination means after the accent delimiter signal is input from the high speech waveform storage means A speech synthesizer characterized by synthesizing speech by extracting a waveform . A speech synthesis program for causing a computer to function as the speech synthesizer according to claim 1.

Images