JPH051494B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is applicable to character strings input by key in information devices such as word processors and personal computers in order to check the correctness of input from the keyboard. This invention relates to a method for synthesizing and outputting speech.

[Prior art and its problems] Recently, information devices such as word processors and personal computers are rapidly becoming popular in offices and households, and the key input method for such information devices is as follows. Methods such as kana input and romaji input have been proposed. Currently, the operator checks whether there are any errors in key input by looking at the character strings displayed on the display screen.

However, in this case, novice operators in particular have to perform input work while shifting their viewpoint between the original, the keyboard, and the display screen, which not only reduces input efficiency but also causes fatigue. The problem is that the degree increases. Further, even in the case of an operator who is relatively familiar with key operations, the above-mentioned problems cannot be avoided, although there may be differences in degree, since the operator must alternately look at both the document and the display surface.

In order to solve the above problem, for example, if the key input speed is faster than the audio output speed, it is possible to stop the audio output of a syllable in the middle, but in this case, a dropped sound may appear at the beginning of the next syllable. There was a problem that the intelligibility deteriorated due to mixing.

[Object of the Invention] The object of the present invention has been made in view of the above-mentioned problems, and it is possible to easily confirm the correctness of key input and eliminate the above-mentioned click sound. An object of the present invention is to provide an audio output method for confirming key input.

[Structure of the Invention] In order to achieve the above object, the present invention stores audio data for each sound (50 sounds, voiced sounds, etc.) in advance, and uses the audio data to generate sounds corresponding to characters input by key. It synthesizes and outputs each syllable (one unit of speech such as "a", "ka", etc.) based on The basic feature is that the audio output time of each syllable is made variable so that the duration of occurrence of the syllable is shortened.

The equipment to realize this type of confirmation method is
means for storing voice data; means for capturing key input; means for detecting the speed of key input; The device includes means for synthesizing speech corresponding to key input based on data, and means for outputting the synthesized speech.

Here, as a method of changing the audio output time of each syllable, for example, in the audio output of each syllable, the audio output speed (the reciprocal of the total time of the utterance duration of a single syllable and the silent part time) is constant, and the key input A method is adopted in which the audio output of the syllable is terminated when the audio output time corresponding to the speed has elapsed. As a result, the audio output time of each syllable is adjusted to the standard output time (1
The time required to output all syllables worth of audio data at a predetermined speed) is reduced by an arbitrary reduction rate depending on the key input speed.

In this case, if the key input speed is relatively fast and there is still audio data to synthesize and output a certain syllable after the audio output time for that syllable has passed, the remaining audio data will be ignored. The audio output is interrupted midway. On the other hand, if the key input speed is sufficiently slow, the audio output of each syllable will not be interrupted midway and will continue throughout the reference output time.

Furthermore, in the present invention, when the speed of key input is faster than the speed of audio output, the audio output is near the end of the audio output time and the audio output is near the zero point of the peak-to-peak output amplitude range. The configuration is such that when the syllable reaches the end of the syllable, the audio output of the syllable is terminated. This makes it possible to eliminate the drop sound that occurs at the beginning of the next syllable.

Note that when outputting audio using the method described above, it is preferable that the silent time inserted between syllables and the duration of audio output of one syllable be set within the following ranges.

() Silent time between syllables (voice interval) When outputting speech in units of syllables, unless a certain amount of utterance interval is taken, the clarity of the output voice will deteriorate. This utterance interval is, for example, about 50 to 100 ms.

() Speech output time of one syllable When a person utters a single syllable discretely, the duration of the utterance is approximately 300 ms on average for 100 Japanese syllables at normal speed. Therefore, the reference output time can be set to about 300 ms, for example. In that case, the average key input speed is 2.5 syllables/s
If it is a degree, even if the above-mentioned utterance interval is considered,
Key input can be sufficiently followed without making the duration of audio output of each syllable shorter than the reference output time.

On the other hand, if key input is performed at high speed, the audio output time of one syllable must be shorter than the standard output time. In that case, the upper limit of the key input speed is usually estimated to be about 5 syllables/s, and in order to correspond to this expected maximum key input speed, it is necessary to shorten the output time of one syllable to about 150 ms. Note that this value of 150ms is considered to be close to the limit in terms of intelligibility, so the length of one syllable should be the minimum.
It is preferable to set it to about 150ms.

[Effects of the Invention] As described above, in the present invention, since the voice corresponding to the characters inputted by the keys is synthesized and output, the operator can check whether there are any errors in the key inputs by checking the display screen. You can easily check the surface based on your hearing without having to look at the surface. As a result, it is possible to improve the efficiency of key input and to reduce the degree of fatigue.

Furthermore, since the audio output time of each syllable is made variable and the audio output time becomes shorter as the key input speed increases, the delay time of audio output with respect to key input can be minimized. Thereby, the operator's discomfort can be reduced.

Further, as described above, when the audio output time is set according to the speed of key input, and the speed of the key input is faster than the speed of the audio output, near the end of the audio output time, Since the audio output of the syllable is discontinued when the audio output reaches the zero point of the peak-to-peak output amplitude range, essentially only the audio output time of the vowel part of each syllable changes. In this case, the consonant that forms the beginning of a word is always output in the same way, maintaining the clarity of the output voice, and also eliminating the above-mentioned dropped consonant, which makes it possible to produce the next syllable. This has the advantage that the voice is clearly output.

[Example] Hereinafter, an example of the present invention will be described with reference to the accompanying drawings.

FIG. 2 shows a block configuration of an information device that employs the confirmation method according to the present invention, and the information device includes a keyboard 1 and an information device such as a personal computer (PC) or a word processor (WP). It basically consists of a main body 2 and a speech synthesis output section 3.

The keyboard 1 has a key input processing CPU 4,
It is equipped with a key matrix 5 in which key switches corresponding to each key are arranged vertically and horizontally, and is used for key input processing.
The CPU 4 cyclically outputs a strobe to the key matrix 5 to check whether each key has been pressed.

In addition, the key input processing CPU 4 transmits strobe signals 7, 7 indicating the presence or absence of key input to the main body 2 and the synthesis processing CPU 6 included in the speech synthesis output section 3, and also sends strobe signals 7, 7 indicating the presence or absence of key input. When the key switch 4 is pressed, the corresponding character codes 8, 8 are sent to the main body 2 and the composition processing CPU 6. The subsequent processing within the main body 2 is not related to the gist of the present invention, and therefore will not be described here.

The synthesis processing CPU 6 is connected to a program ROM 11, a RAM 12, a code ROM 13, a timer 14, and a D/A converter 15 via a bus 10 for addressing, data transfer, and control, and by accessing these elements, Key input processing
The character code string supplied from the CPU 4 is converted into syllables, and speech synthesis processing is performed for each syllable. In addition, each of the above-mentioned elements has the following functions, respectively.

() Program ROM11...Composition processing CPU6
Contains programs for operating the.

() RAM 12...Buffer 1 that temporarily stores the character code supplied from the key input processing CPU 4, and the syllable code (which specifies each sound such as "a" and "ka") corresponding to the character code or character code string. ) consists of a buffer 2 that temporarily stores the data. Both buffers 1 and 2 are FIFOs (first-in-first-input) that process each code in the order in which they are stored.
out) type.

() Data ROM 13: Audio data corresponding to each syllable code is stored. The audio data is usually PCM (pulse code modulation) code or
It is given in a waveform encoding format such as ADPCM (differential pulse code modulation) code. In other words, 50
Data obtained by sampling a waveform representing each sound such as a sound at predetermined sampling intervals and quantizing each sampling value is stored as audio data.

() Timer 14...Speech synthesis processing is performed by an interrupt (described later), and this interrupt requests an interrupt by outputting a pulse every predetermined time.

() D/A converter 15: converts the digital waveform representing each syllable synthesized by the synthesis processing CPU 6 into an analog waveform based on the audio data in the data ROM 13. This analog waveform is sent to a speaker 17 via an output section 16 that includes a bottom pass filter (LPF) and an amplifier (AMP), and is converted into an acoustic signal by the speaker 17 and output as an audio signal.

The general flow of the key input confirmation process will be described below with reference to the flowchart shown in FIG.

() In step S1, it is determined whether or not there is key data (key input) based on the strobe signal 7. If there is key input, the process proceeds to S2 and the character code supplied from the key input processing CPU 4 is determined. is temporarily stored in buffer 1.

() Subsequently, in S3 and S4, syllable analysis of the character code in buffer 1 is performed to determine whether the character code or character code string forms a valid syllable. If a valid syllable has not yet been formed (for example, only consonants have been input in Roman alphabet input), return to S1; if a valid syllable has been formed, proceed to S5 and respond to that syllable. The syllable code is stored in buffer 2.

() If there is no key data in S1, the process proceeds to S6, where it is determined whether or not the voice corresponding to the characters that have already been input is being synthesized. If yes,
Returning to S1, if no, it is further determined in S7 whether or not Buffer 2 is a syllable code, that is, whether there is a syllable that has already been keyed in but has not yet been outputted as a voice. .

() As a result of the determination, if there is no syllable code in buffer 2, the process returns to S1, and if so, proceeds to S8, where the key input speed is estimated from the number of syllable codes in buffer 2, and based on the key input speed. , the reduction rate of the audio output time (ratio of the actual output time to the reference output time) is calculated. In that case, if the number of codes in buffer 2 is large, it is assumed that the key input speed is high, and by setting a large reduction rate, the audio output time per syllable is shortened.
If the number of codes in the syllable is small, it is assumed that the key input speed is low, and the shortening rate is set small, thereby lengthening the audio output time per syllable. In this way, there is no need to detect the actual key input speed, so the circuit configuration can be simplified. However, it is also possible to detect the actual key input speed.

() Continuing in S9, the synthesis control data corresponding to the syllable code located at the beginning of Buffer 2 (consisting of the start address and number of data of the audio data corresponding to the first code in Buffer 2...Other synthesis control data is the same) ) in the register and counter, give interrupt permission in S10, and return to S1.

Next, a subroutine for speech synthesis will be explained with reference to FIG.

This speech synthesis subroutine is processed by an interrupt. i.e. main routine S10
Alternatively, when interrupt permission is granted in S111 of this subroutine, the timer 14 outputs a pulse for an interrupt request, thereby starting the speech synthesis subroutine. The above-mentioned pulse for interrupt request is outputted every reciprocal time of the sampling frequency of the voice to be synthesized. For example, if the sampling frequency is 10KHz, the time interval between pulses will be 100μs.

When the speech synthesis subroutine is started, () In S101, the registers used so far are stored, and in S102, one-syllable speech synthesis processing is performed based on the synthesis control data set in S9 of the main routine. Determine whether or not the process has ended. If the process has ended, the registers are restored in S103 and the process returns to the main routine.

() If the speech synthesis output process for one syllable has not been completed in S102, the process proceeds to S104, where one sample of speech data of the syllable to be output is read from the data ROM 13 based on the synthesis control data, and the D/ The data is sent to the A conversion unit 15 and one sample is synthesized. Further, in S105, the number of data set in the counter is decremented, and the address in the data ROM 13 to be read is increased, in preparation for reading the next sample.

() In S106, it is determined whether the end of the audio output of one syllable is approaching, and if it is not near the end, S107 to S109 are skipped and the process proceeds to S110, the registers are restored, and further interrupt permission is granted. After (S111), the process returns to the main routine.

() In S106, if it is near the end of the audio output of one syllable, proceed to S107, and check whether the output level of the sample is near the midpoint of the output range, in other words, near the zero point of the full range of the D/A output. It is determined whether or not. If it is not near the midpoint, wait until it is near the midpoint, and if it is near the midpoint, execute S108
After proceeding to , outputting the midpoint value, and further setting "φ" to the data number of the counter in S109, and completing the one-syllable speech synthesis output process, proceed to S110.
In S111, the same process as described above is performed and the process returns to the main routine. In this way, by making sure that the audio output of each syllable is terminated at the midpoint, it is possible to prevent the above-mentioned dropout from occurring.

Note that if the speech synthesis process is performed using an interrupt as described above, the main routine processing can be performed except for the processing time required for each sample in the speech synthesis of each syllable, so key input processing is not required. Including,
Syllable analysis and speech synthesis can be performed in parallel.

Finally, a specific example of speech synthesis will be explained with reference to the time chart shown in FIG.

That is, in Fig. 1, a indicates the timing of key input, and b indicates the timing of audio output of each syllable when the audio output time is not changed (when audio output is performed at the standard output time for all syllables). ,c
shows the timing of audio output of each syllable when the audio output time is changed (shortened) according to the key input speed. As is clear from the figure, when the audio output time is shortened in accordance with the key input speed, the delay time of audio output with respect to key input is considerably reduced.

In the above embodiment, the voice synthesis output section 3 is configured separately from the keyboard 1, but it goes without saying that the voice synthesis output section 3 can also be built into the keyboard 1.

[Brief explanation of drawings]

1a, b, and c are time charts showing specific examples of the confirmation method according to the present invention, FIG. 2 is a block configuration diagram of an information device adopting the confirmation method according to the present invention, and FIG. FIG. 4 is a flowchart showing the general flow of a confirmation processing routine using the confirmation method according to the present invention. FIG. 4 is a flowchart showing a speech synthesis subroutine that is inserted into the routine of FIG. 1...Keyboard (key input unit), 2...Information equipment main body, 3...Speech synthesis output unit, 6...Synthesis processing CPU, 11...Program ROM, 12...
RAM, 13...Data ROM (audio data storage unit), 14...Timer, 15...D/A conversion unit,
16...Output unit, 17...Speaker (audio output unit).

Claims (1)

[Claims] 1.Speech data for each sound is stored in advance, and the sound corresponding to the character entered by the key is synthesized syllable by syllable based on the above sound data and output at a constant sound output speed, and at the same time as the speed of the key input. Set the audio output time accordingly, and if the speed of the key input is faster than the speed of the audio output, the output amplitude range is near the end of the audio output time and the audio output is peak-to-peak. An audio output method for confirming key input is characterized in that the audio output of the syllable is stopped when the syllable reaches a zero point.