JP6790959B2 - Speech synthesizer, speech synthesis method and speech synthesis system, and computer program for speech synthesis - Google Patents

Description

The present invention relates to, for example, a speech synthesizer for generating a speech string used for generating a synthetic speech signal from text data, a speech synthesis method and a speech synthesis system, and a computer program for speech synthesis.

A speech synthesis technique for reading input text data and automatically synthesizing a synthetic speech signal representing a speech waveform corresponding to the text data is being studied. This speech synthesis technique is used in, for example, a so-called reading application that outputs text data such as a news article posted on a website or the like as a synthesized speech (hereinafter, simply referred to as a synthetic speech).

In reading the text data, it is important that the text data can be read accurately and that the synthetic voice is given an appropriate accent in order to reproduce the synthetic voice that does not give a sense of discomfort to the user. Therefore, when the text data with ruby is input and the reading of the ruby equivalent character string and the reading of the immediately preceding character string do not match, the text data composite sound including the immediately preceding character string and the ruby equivalent character string is input. A technique for generating has been proposed (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2005-309327

In Japanese, the position of an accent in a compound word formed from a plurality of morphemes may change from the position of each accent of those morphemes. Such a change in the position of an accent is called an accent combination. Since such accent combination is not assumed in the prior art, the position of the accent given to the synthetic speech signal may be inappropriate when the compound word contains a character string with ruby characters. ..

As one aspect, it is an object of the present invention to provide a speech synthesizer capable of giving an appropriate accent to a synthetic speech generated for text data including a phrase in which ruby is added.

According to one embodiment, a speech synthesizer is provided. This voice synthesizer executes a morpheme analysis on text data including words and phrases with ruby, and divides the text data into morpheme combinations, and a morpheme combination included in the morpheme combination. Of these, the extraction unit that extracts the morphemes contained in the phrase with ruby, and whether or not to form a compound word with the morphemes that are continuous in front or behind the morpheme of the morpheme contained in the phrase with ruby. It has a combination determination unit for determining whether or not, and an accent addition unit for giving an accent to the compound word as a whole when the morphemes included in the phrase to which ruby is added form a compound word.

Appropriate accents can be added to the synthetic speech generated for text data containing words with ruby characters.

It is a schematic block diagram of a speech synthesizer. It is a functional block diagram of the processing part which a voice synthesizer has. It is a figure which shows an example of the morphological analysis. It is a figure which shows an example of the compound word registered in a language dictionary. It is a figure which shows an example of the reference table used for registering a compound word in a language dictionary. It is an operation flowchart of an accent addition process. It is an operation flowchart of the voice synthesis processing. It is a figure which shows an example of the processing result by this embodiment. It is a schematic block diagram of a client-server system in which a speech synthesizer is mounted according to a modification.

Hereinafter, the speech synthesizer will be described with reference to the drawings.
This speech synthesizer performs morphological analysis on the input text data including the words with ruby characters, and extracts the morphemes with ruby characters. Then, this speech synthesizer determines whether or not a compound word is formed by the extracted morpheme with ruby and the morpheme before and after the extracted morpheme, and if the compound word is formed, the compound word is referred to from the language dictionary. Search for and use compound word accents. Further, when the compound word including the morpheme with ruby is not formed, this speech synthesizer searches the language dictionary for the accent corresponding to the morpheme with ruby. Thereby, this speech synthesizer can give an appropriate accent to the synthetic speech signal generated for the text data including the phrase with ruby.

FIG. 1 is a schematic configuration diagram of a speech synthesizer according to one embodiment. In the present embodiment, the voice synthesizer 1 has an operation unit 2, a display unit 3, a communication interface unit 4, a storage unit 5, a processing unit 6, and a voice output unit 7. The voice synthesizer 1 can be, for example, a mobile device such as a mobile phone or a so-called tablet computer, a desktop computer, a web server, or the like.

The operation unit 2 has, for example, a keyboard and a pointing device such as a mouse. Then, the operation unit 2 specifies, for example, the text data to be voice-synthesized from the text data stored in the storage unit 5 according to the operation by the user, and sets a signal for designating the designated text data. Pass it to the processing unit 6. In the present embodiment, the text data to be voice-synthesized includes words with ruby characters. However, even when the text data to be voice-synthesized does not include words with ruby, the voice synthesizer can generate a synthetic voice signal for the text data.

The display unit 3 has a display device such as a liquid crystal display. Then, the display unit 3 displays the detected synonyms and the like. The operation unit 2 and the display unit 3 may be integrally formed as a touch panel display.

The communication interface unit 4 has an interface circuit for connecting the voice synthesizer 1 to the communication network, and acquires various information via the communication network. For example, the communication interface unit 4 acquires text data including words with ruby characters, which is a voice synthesis target, from another device connected to the voice synthesis device 1 via a communication network.

Further, the communication interface unit 4 outputs the synthesized voice signal generated for the text data to be voice synthesis received from the processing unit 6 to another device connected to the voice synthesis device 1 via the communication network. You may. Alternatively, the communication interface unit 4 is connected to the voice synthesizer 1 via a communication network with a phonetic character string to which information for generating a synthetic voice signal such as a reading and an accent position with respect to the text data is added. It may be output to other devices.

The storage unit 5 has, for example, a semiconductor memory circuit such as a non-volatile read-only semiconductor memory and a volatile read / write semiconductor memory. The storage unit 5 may further include a magnetic storage device or an optical storage device. Then, the storage unit 5 stores various computer programs used in the processing unit 6 and various data used in the voice synthesis processing.
The storage unit 5 generates data used in the speech synthesis processing, for example, a reading of each morpheme and each compound word, a language dictionary representing accent and part-speech information, a connection rule representing the connection cost between morphemes, and compound word generation. Memorize rules and so on.

The processing unit 6 includes one or more processors, a memory circuit, and peripheral circuits. Then, the processing unit 6 executes the voice synthesis processing on the text data designated via the operation unit 2. The details of the speech synthesis process will be described later.

The voice output unit 7 has, for example, a speaker, and outputs a synthetic voice signal generated for text data as voice (that is, synthetic voice).

FIG. 2 is a functional block diagram of the processing unit 6. The processing unit 6 includes a text analysis unit 11, an extraction unit 12, a combination determination unit 13, an accent addition unit 14, and a speech synthesis unit 15.
Each of these units included in the processing unit 6 is, for example, a functional module realized by a computer program running on the processor included in the processing unit 6. Alternatively, each of these units included in the processing unit 6 may be mounted on the speech synthesizer 1 separately from the processor included in the processing unit 6 as one or a plurality of integrated circuits that realize the functions of the respective units.

The text analysis unit 11 executes morphological analysis on text data including words with ruby characters, which is a voice synthesis target and is designated via the operation unit 2, and converts the text data into a combination of morphemes. To divide. The text data to be voice-synthesized can be, for example, data according to any of various formats in which a character string including ruby can be described.

In the present embodiment, the text analysis unit 11 performs morphological analysis on the text data by referring to, for example, a language dictionary, and divides the text data into morphemes for all possible morphemes. Find the combination. Then, the text analysis unit 11 calculates an evaluation value indicating the certainty of the combination for each combination of morphemes. At that time, the text analysis unit 11 refers to the connection rule applied between the continuous morphemes for each of the continuous morphemes included in the combination of the morphemes of interest, and refers to the connection rules between the continuous morphemes. Find the cost. Then, the text analysis unit 11 sets the total cost obtained for each of the consecutive morphemes included in the combination of the morphemes of interest as the evaluation value for the combination of the morphemes of interest. Then, the text analysis unit 11 selects the combination having the smallest evaluation value among the obtained combinations of morphemes as the morphological analysis result for the text data. The text analysis unit 11 reduces the amount of calculation required to calculate the evaluation value by using a method based on a dynamic programming method such as the Viterbi algorithm in order to specify the combination of morphemes that minimizes the evaluation value. May be good.

FIG. 3 is a diagram showing an example of morphological analysis. In this example, the text data 300 "Today is sunny" is input. By performing morphological analysis on the text data 300 with reference to the language dictionary 301, a lattice structure 302 representing a possible combination of morphemes can be obtained. In this lattice structure 302, the individual blocks 303 represent one morpheme, and the individual arrows 304 represent possible connections between the morphemes. Then, an evaluation value is calculated for each combination of morphemes, and as a combination of morphemes whose evaluation value is the minimum, a set of "today" (noun), ha (particle), and "sunny" (noun) is obtained.

The text analysis unit 11 may divide the text data to be voice-synthesized into combinations of morphemes according to any of various morphological analysis methods other than the above.

The text analysis unit 11 notifies the extraction unit 12, the combination determination unit 13, and the accenting unit 14 of the reading and part of speech of each morpheme included in the specified combination of morphemes and the order of arrangement of each morpheme.

The extraction unit 12 extracts morphemes corresponding to ruby in the combination of morphemes specified for the text data. For example, the extraction unit 12 determines, for each of the morphemes included in the specified combination of morphemes, whether or not ruby is added to the words of the original text data including the morphemes. Then, the extraction unit 12 sets the morpheme as a morpheme corresponding to the ruby if the word containing the morpheme is given ruby, and sets the morpheme as the ruby if the word containing the morpheme is not given ruby. It is a morpheme that does not correspond to.
The extraction unit 12 notifies the combination determination unit 13 of each of the morphemes corresponding to ruby.

The combination determination unit 13 determines whether or not each of the morphemes corresponding to ruby is included in the compound word. Therefore, in the combination determination unit 13, the relationship between the part of speech of the morpheme corresponding to ruby and the part of speech of the morpheme before or after the morpheme is expressed in the compound word generation rule, and the relationship of the part of speech that becomes a compound word. Judge whether or not the condition is satisfied. Then, when the relationship between the part of speech of the morpheme corresponding to rubi and the part of speech of the morpheme before or after it satisfies the relationship of the part of speech that is a compound word, the combination determination unit 13 sets the morpheme corresponding to rubi and the morpheme before or after it. The latter morpheme is regarded as one compound word. On the other hand, if the relationship between the part of speech of the morpheme corresponding to ruby and the part of speech of any of the morphemes before and after it does not satisfy the relationship of the part of speech that is a compound word, the combination determination unit 13 determines that the morpheme corresponding to ruby is It is determined that a compound word is not formed.

The relationship between the part of speech that is a compound word expressed in the compound word generation rule can be, for example, any of a prefix + noun, a noun + suffix, and a noun + noun.

The combination determination unit 13 sends the result of determining whether or not to form a compound word for each of the morphemes corresponding to the ruby and the compound word including the morpheme for the morpheme forming the compound word to the accent giving unit 14. Notice.

The accent giving unit 14 gives an accent to each morpheme included in the combination of morphemes specified for the text data. At that time, when the morpheme corresponding to the ruby forms a compound word and the compound word is registered in the language dictionary, the accent adding unit 14 adds the entire compound word to the language dictionary. Gives an accent to the registered compound word. For example, the accent adding unit 14 searches the language dictionary for a compound word having a notation and reading that matches the reading specified by referring to the notation and ruby of the compound word of interest, and corresponds to the matching notation and reading. An accent may be added to the compound word of interest.

FIG. 4 is a diagram showing an example of compound words registered in a language dictionary. The linguistic dictionary 400 includes, for each notation of a compound word, a phonetic character string representing a reading and an accent position corresponding to the notation, and a part of speech of the compound word. For example, the compound word "Nihonbashi Station" is associated with two phonetic strings "Nihonbashi'Eki" and "Nipponbashi' Eki", and in both readings, the sound of "Shi" is used. It is shown to have an accent. The position of the accent is indicated by the symbol "'". It is also shown that the part of speech of the compound word "Nihonbashi Station" is a noun. Therefore, for example, it is assumed that the compound word "Nihonbashi Station" is included in the combination of morphemes specified for the text data, and the reading is "Nipponbashieki". In this case, the accent giving unit 14 can give an accent to the sound of "shi" for the compound word "Nihonbashi station" by referring to the language dictionary 400.

The reading and accent of the compound word are determined in advance for each morpheme capable of forming the compound word by using a reference table representing the accent combination rules applied to the morpheme, and are registered in the language dictionary. The compound word may be additionally registered in the language dictionary in which the morpheme other than the compound word is registered, or the compound word is registered separately from the language dictionary in which the morpheme other than the compound word is registered. A language dictionary may be prepared and stored in the storage unit 5.

FIG. 5 is a diagram showing an example of a reference table used for registering a compound word in a language dictionary. The reference table 500 represents the notation, reading, accent position, part of speech, and accent combination style of each morpheme that can form a compound word. For example, regarding the morpheme "procedure", it is shown that the reading is "Tetsuzuki", the accent is "Tsu", the part of speech is a noun, and the combination mode is C1.

In this example, as the combination mode, Non-Patent Document 1 (Yoshinori Sagisaka, Yamato Sato, "Accent Rule of Japanese Word Chain", IEICE Transactions D, Vol.J66-D, No.7, pp With reference to .849-856, July 1983), five types of binding modes C1 to C5 are shown. Hereinafter, the preceding morpheme will be described as a “preceding word” and the rear morpheme will be described as a “successor word” in the combination.
The combination mode C1 indicates that the accent of the preceding word disappears. For example, when the compound word "moving procedure" is formed by combining the morpheme "moving (tenkyo)" and the morpheme "procedure (tetsu'zuki)", the reading and accent of the compound word is "tenkyo tetsu'zuki". ".

The combination mode C2 indicates that the accent of the preceding word is lost and the accent is located at the beginning of the following word. For example, when the compound word "weather forecast" is formed by combining the morpheme "weather" and the morpheme "forecast (yoho)", the reading and accent of the compound word is "tenkiyo'ho". Become.

The combination mode C3 indicates that the accent is located at the end of the preceding word. For example, when the compound word "Shinagawa Station" is formed by combining the morpheme "Shinagawa" and the morpheme "Station (Eki)", the reading and accent of the compound word is "Shinagawa'Eki".

The combination mode C4 indicates that there is no accent. For example, when the compound word "Rishiri Island" is formed by combining the morpheme "Rishiri" and the morpheme "Island (To)", the reading and accent of the compound word is "Rishiri To".

The combination mode C5 indicates that the accent of the following word is lost. For example, when the compound word "Satoden" is formed by combining the morpheme "Sato" and the morpheme "Do'no", the reading and accent of the compound word is "Satodono". ".

FIG. 6 is an operation flowchart of the accent giving process executed by the accent giving unit 14. The accent giving unit 14 gives an accent to each morpheme included in the combination of morphemes specified for the text data according to the following operation flowchart.

The accent giving unit 14 determines whether or not the morpheme of interest is a morpheme included in the phrase to which ruby is added (step S101). If the morpheme of interest is not included in the phrase to which ruby is added (step S101-No), the accenting unit 14 imparts the accent of the morpheme registered in the language dictionary to the morpheme of interest (step S102). ).

On the other hand, if the morpheme of interest is included in the phrase to which ruby is added (step S101-Yes), the accent adding unit 14 determines whether or not the morpheme of interest is included in the compound word (step S103). .. If the morpheme of interest is included in the compound word (step S103-Yes), the accent adding unit 14 determines whether or not the compound word including the morpheme of interest is registered in the language dictionary (step S104).

If the compound word containing the morpheme of interest is registered in the language dictionary (step S104-Yes), the accenting unit 14 refers to the language dictionary and makes the compound word the compound registered in the language dictionary. Add a word accent (step S105). On the other hand, if the compound word containing the morpheme of interest is not registered in the language dictionary (step S104-No), the accenting unit 14 has another word having a reading and part of speech that matches the reading and part of speech of the compound word. It is determined whether or not it is registered in the language dictionary (step S106). When another word having a reading and part of speech matching the reading and part of speech of the compound word is registered in the language dictionary (step S106-Yes), the accenting unit 14 matches the reading and part of speech of the compound word. Identify the accents of other words that have readings and parts of speech. Then, the accent giving unit 14 gives the specified accent to the compound word (step S107). When a plurality of such other words are registered in the language dictionary, the accent adding unit 14 assigns the most frequent accent among the accents of the plurality of other words to the compound word. May be good.

On the other hand, when another word having a reading and part of speech that matches the reading and part of speech of the compound word is not registered in the language dictionary (step S106-No), the accent giving unit 14 registers the morpheme of interest in the language dictionary. It is determined whether or not this is done (step S108). Further, in step S103, even when the morpheme of interest is not included in the compound word (step S103-No), the accent giving unit 14 determines whether or not the morpheme of interest is registered in the language dictionary (step S108). ).

If the morpheme of interest is registered in the language dictionary (step S108-Yes), the accent adding unit 14 imparts the accent of the morpheme registered in the language dictionary to the morpheme of interest (step S109). On the other hand, if the morpheme of interest is not registered in the language dictionary (step S108-No), the accent adding unit 14 gives a default accent to the morpheme of interest (step S110). The default accent can be, for example, no accent.

After steps S102, S105, S107, S109 or S110, the accenting unit 14 ends the accenting process. According to the modified example, the processing of steps S106 and S107 may be omitted. In this case, if the compound word is not registered in the language dictionary in step S104, the accent giving unit 14 may execute the processes after step S108.

The accenting unit 14 gives an accent to each morpheme included in the combination of morphemes specified for the text data. Then, the accent giving unit 14 creates a phonetic character string representing the reading and accent position of the entire text data by arranging the phonetic character strings representing the reading and accent positions of each morpheme according to the order of the morphemes. .. Then, the accent giving unit 14 outputs the phonetic character string of the text data to the voice synthesis unit 15.

The voice synthesis unit 15 generates a synthetic voice signal representing the voice waveform of the text data based on the phonetic character string of the text data.
The voice synthesis unit 15 may generate a synthetic voice signal of text data according to any of various voice synthesis methods. For example, the speech synthesis unit 15 generates a target prosody for generating a synthetic speech signal based on a phonetic character string. Therefore, the speech synthesis unit 15 reads a plurality of prosodic models from the storage unit 5. In this prosody model, the position where the voice is raised and the position where the voice is lowered are expressed in chronological order. Then, the voice synthesis unit 15 selects the prosody model that most closely matches the position of the accent shown in the phonetic character string among the plurality of prosody models. Then, the voice synthesis unit 15 determines the position where the voice becomes high or low and the voice intonation with respect to the phonetic character string according to the selected prosodic model and various parameters referred to for generating the synthetic voice signal. , Pitch, etc. to create the target prosody. The target prosody includes the phoneme length and pitch frequency for each phoneme that is the unit that determines the speech waveform. The phoneme can be, for example, one vowel or one consonant.

The voice synthesis unit 15 creates a synthetic voice signal according to the generated target prosody, for example, by an HMM (Hidden Markov Model) synthesis method, a phoneme connection method, or a corpus-based method.
For example, the voice synthesis unit 15 selects a voice waveform closest to the phonetic length and pitch frequency of the target prosody for each voice element from a plurality of voice waveforms registered in the voice waveform dictionary by pattern matching, for example. Therefore, the voice synthesis unit 15 reads the voice waveform dictionary from the storage unit 5. The voice waveform dictionary records a plurality of voice waveforms and identification numbers of each voice waveform. The voice waveform is, for example, a waveform signal in phoneme units.
Further, the speech synthesizer 15 connects the selected speech waveforms with the corresponding phoneme waveform patterns shown in the target prosody so that the speech waveforms selected for each phoneme can be connected along the target prosody. The amount of deviation may be calculated as waveform conversion information.
The voice synthesis unit 15 creates waveform generation information including an identification number of a voice waveform selected for each phoneme. The waveform generation information may further include waveform conversion information.

The voice synthesis unit 15 reads the voice waveform signal corresponding to the identification number of the voice waveform of each phoneme included in the waveform generation information from the storage unit 5. Then, the voice synthesis unit 15 creates a synthetic voice signal by continuously connecting each voice waveform signal. When the waveform generation information includes waveform conversion information, the voice synthesis unit 15 corrects each voice waveform signal according to the waveform conversion information obtained for the corresponding phonetic element, and continuously connects the voice waveform signals. By doing so, a synthetic voice signal may be created.

The voice synthesis unit 15 outputs the generated synthetic voice signal to the voice output unit 7. Alternatively, the voice synthesis unit 15 may store the generated synthetic voice signal in the storage unit 5 or output it to another device via the communication interface unit 4.

FIG. 7 is an operation flowchart of the voice synthesis process.
The text analysis unit 11 performs morphological analysis on the text data including the words and phrases with ruby characters designated via the operation unit 2, and divides the text data into combinations of morphemes (step S201). Then, the extraction unit 12 extracts the morpheme contained in the phrase to which ruby is added from the morpheme contained in the combination of morphemes (step S202).

The combination determination unit 13 identifies the morpheme included in the compound word among the morphemes included in the phrase to which ruby is added (step S203). The accent giving unit 14 creates a phonetic character string of text data by giving an accent to each morpheme (step S204). Then, the voice synthesis unit 15 generates a synthetic voice signal of text data based on the phonetic character string (step S205). Then, the processing unit 6 ends the speech synthesis process.

FIG. 8 is a diagram showing an example of the processing result according to the present embodiment. In this example, the input text data 800 includes the words "Professional baseball (Yakyu) ha" and "Nihonbashi (Nipponbashi) station" with ruby characters. The readings shown in parentheses represent ruby. These words and phrases are divided into "professional / baseball / ha" and "Nihonbashi / station / no", respectively, by morphological analysis. Of these, the morphemes "baseball" and "Nihonbashi" are extracted as morphemes corresponding to ruby, and these morphemes are determined to be combined when they form the compound words "professional baseball" and "Nihonbashi station", respectively. Then, the accent "Proya'cue" of the compound word "Professional baseball" registered in the language dictionary 801 is given to the compound word "Professional baseball" including the morpheme corresponding to the word with the ruby in the text data 800. Will be done. Similarly, the accent "Nipponbashi'Eki" of the compound word "Nihonbashi Station" registered in the language dictionary 801 contains the compound word "Nihonbashi Station" containing the morpheme corresponding to the word with the ruby in the text data 800. Is given to. In this example, in the language dictionary 801, two readings and accents "Nihonbashi'Eki" and "Nipponbashi' Eki" are registered for the compound word "Nihonbashi Station". However, in the text data 800, since ruby (Nippombashi) is assigned to the phrase "Nihonbashi Station", the reading and accent "Nippombashi'Eki" corresponding to the ruby is selected.

As described above, this speech synthesizer determines whether or not a compound word is formed with the morphemes before and after the morpheme contained in the phrase to which the ruby is added, and when the compound word is formed, the compound word is formed. Add accents on a word-by-word basis. Therefore, this speech synthesizer can give an appropriate accent to the morpheme contained in the phrase to which ruby is added, even if an accent combination occurs due to the formation of a compound word. As a result, the speech synthesizer can generate a synthetic speech signal that sounds natural to the user.

According to the modification, the voice synthesizer may output a phonetic character string of text data to be voice-synthesized to another device via the communication interface unit 4. Then, in another device, a synthetic voice signal of the text data may be generated based on the phonetic character string. In this case, the voice synthesis unit 15 may be omitted.

Further, according to another modification, the combination determination unit 13 may determine whether or not a compound word is formed together with the morphemes before and after the morpheme in which ruby is not added. Then, when the compound word including the morpheme to which the ruby is not added is registered in the language dictionary, the accent adding unit 14 sets the compound word registered in the language dictionary for the compound word. Word accents may be added.

According to still another variant, the speech synthesizer may be implemented in a client-server system.

FIG. 9 is a schematic configuration diagram of a client-server system in which a voice synthesizer is mounted according to this modified example. The client-server system 100 has a terminal 110 and a server 120, and the terminal 110 and the server 120 can communicate with each other via the communication network 130. There may be a plurality of terminals 110 included in the client-server system 100. Similarly, there may be a plurality of servers 120 included in the client-server system 100.

The terminal 110 has an input unit 111, a storage unit 112, a communication unit 113, a control unit 114, and an audio output unit 115. The input unit 111, the storage unit 112, the communication unit 113, and the audio output unit 115 are connected to the control unit 114 via a bus.

The input unit 111 has an input device for inputting text data, such as a keyboard, and acquires text data including words and phrases with ruby text, which is a target of voice synthesis.
The storage unit 112 has, for example, a non-volatile semiconductor memory and a volatile semiconductor memory. Then, the storage unit 112 stores a computer program for controlling the terminal 110, identification information of the terminal 110, and the like.

The communication unit 113 has an interface circuit for connecting the terminal 110 to the communication network 130. Then, the communication unit 113 transmits the text data received from the control unit 114 to the server 120 via the communication network 130. Further, the communication unit 113 is another example of the input unit, and is another device 140 connected to the communication network 130, for example, a Web server, for text data including words with ruby characters, which is a target of voice synthesis. It may be received from and passed to the control unit 114.
Further, the communication unit 113 receives the synthetic voice signal corresponding to the text data from the server 120 and passes it to the control unit 114.

The control unit 114 includes one or more processors and peripheral circuits thereof. Then, the control unit 114 generates text data including words and phrases with ruby characters, which is a target of voice synthesis, and a transmission signal including identification information of the terminal 110. Then, the control unit 114 transmits the transmission signal to the server 120 via the communication unit 113 and the communication network 130.
Further, the control unit 114 outputs a synthetic voice signal received from the server 120 via the communication network 130 and the communication unit 113 as a synthetic voice signal via the voice output unit 115.

The voice output unit 115 has, for example, a speaker, and outputs a synthetic voice signal received from the control unit 114 as a synthetic voice.

The server 120 has a communication unit 121, a storage unit 122, and a processing unit 123. The communication unit 121 and the storage unit 122 are connected to the processing unit 123 via a bus.

The communication unit 121 has an interface circuit for connecting the server 120 to the communication network 130. Then, the communication unit 121 receives the transmission signal including the text data including the ruby-spoken words and the identification information of the terminal 110, which is the target of voice synthesis, from the terminal 110 via the communication network 130, and the processing unit 123. Pass to. Further, the communication unit 121 transmits the synthetic voice signal received from the processing unit 123 to the terminal 110 via the communication network 130 based on the identification information of the terminal 110.

The storage unit 122 has, for example, a non-volatile semiconductor memory and a volatile semiconductor memory. Then, the storage unit 122 stores a computer program for controlling the server 120, a computer program for voice synthesis, various data used in the voice synthesis process, and the like.

The processing unit 123 includes one or more processors and peripheral circuits thereof. Then, the processing unit 123 realizes each function of the processing unit of the speech synthesizer according to each of the above-described embodiments or modifications. That is, the processing unit 123 divides the text data with ruby characters received from the terminal 110 into combinations of morphemes. The processing unit 123 extracts the morpheme included in the phrase to which ruby is added from each morpheme included in the combination of the morphemes, and determines whether or not the extracted morpheme forms a compound word. Then, when the morpheme corresponding to the ruby forms a compound word, the processing unit 123 adds an accent of the compound word registered in the language dictionary to the compound word. Further, the processing unit 123 creates a phonetic character string for the entire text data by adding an accent to each of the other morphemes by referring to the language dictionary. Then, the processing unit 123 generates a synthetic voice signal based on the phonetic character string. Then, the processing unit 123 transmits the generated synthetic voice signal to the terminal 110 via the communication unit 121 and the communication network 130.

According to this embodiment, each terminal 110 can obtain a synthesized voice signal of the text data only by transmitting the text data to be voice-synthesized to the server 120. Therefore, each terminal 110 does not need to store a language dictionary or the like. Further, each terminal 110 can obtain a synthetic voice signal of text data to be voice-synthesized in real time by using the server 120 to generate the synthetic voice signal.

Further, according to the modified example, a plurality of servers 120 may cooperate to execute voice synthesis processing on one text data. In this case, each of the plurality of servers 120 may share and execute each function of the processing unit of the speech synthesizer according to each of the above-described embodiments or modifications. For example, the processing unit 123 of one of the plurality of servers 120 performs the processing of the text analysis unit 11, the extraction unit 12, the combination determination unit 13, and the accenting unit 14 among the functions of the processing unit of the speech synthesizer. Execute to generate a phonetic character string of the target text data. Then, the server transmits the phonetic character string to another server among the plurality of servers 120. The processing unit 123 of the other server executes the processing of the voice synthesis unit 15 among the functions of the processing unit of the voice synthesis device, and generates a synthetic voice signal based on the received phonetic character string. Then, the other server returns the generated synthetic voice signal to the one server. Then, the one server transmits the synthesized voice signal to the terminal 110. Alternatively, another server itself may transmit the generated synthetic voice signal to the terminal 110.

A computer program that enables a computer to realize each function of the processing unit of the speech synthesizer according to each of the above embodiments or modifications may be provided in a form recorded on a recording medium that can be read by a computer. Such a recording medium can be, for example, a magnetic recording medium, an optical recording medium, or a semiconductor memory.

All examples and specific terms given herein are intended for teaching purposes to help the reader understand the invention and the concepts contributed by the inventor to the promotion of the art. There is, and should be construed not to be limited to the constitution of any example herein, such specific examples and conditions relating to exhibiting superiority and inferiority of the present invention. Although embodiments of the present invention have been described in detail, it should be understood that various modifications, substitutions and modifications can be made thereto without departing from the spirit and scope of the invention.

1 Speech synthesizer 2 Operation unit 3 Display unit 4 Communication interface unit 5 Storage unit 6 Processing unit 7 Voice output unit 11 Text analysis unit 12 Extraction unit 13 Coupling judgment unit 14 Accenting unit 15 Voice synthesis unit 100 Client-server system 110 Terminal 120 Server 130 Communication network 111 Input unit 112 Storage unit 113 Communication unit 114 Control unit 115 Voice output unit 121 Communication unit 122 Storage unit 123 Processing unit

Claims (7)

A text analysis unit that divides the text data into combinations of morphemes by performing morphological analysis on the text data containing words with ruby characters.
Among the morphemes contained in the combination of the morphemes, an extraction unit for extracting the morphemes contained in the phrase to which the ruby is added, and an extraction unit.
A combination determination unit that determines whether or not a compound word is formed together with a morpheme that is continuous on the front side or the back side of the morpheme of the morpheme included in the phrase to which the ruby is added.
When the morpheme contained in the phrase to which the ruby is added forms the compound word, an accent giving portion that gives an accent to the compound word as a whole and an accent adding portion.
A voice synthesizer having. The accent adding unit generates a phonetic character string representing the reading of the entire text data and the position of the accent by giving an accent to each morpheme included in the combination of the morphemes.
The voice synthesizer according to claim 1, further comprising a voice synthesizer that generates a synthetic voice signal representing a voice waveform of the text data based on the voice string. It also has a storage unit that stores a linguistic dictionary that represents readings and accents for each morpheme and each compound word.
When the morpheme contained in the phrase to which the ruby is added forms the compound word and the compound word is registered in the language dictionary, the accent adding portion is registered in the language dictionary in the compound word. When the morpheme contained in the phrase to which the ruby is added does not form the compound word and the morpheme is registered in the language dictionary, the morpheme is added. The voice synthesizer according to claim 1 or 2, which imparts an accent of the morpheme registered in the language dictionary to the morpheme. The accent giving portion is registered in the language dictionary when the morpheme contained in the phrase to which the ruby is added forms the compound word and the compound word is not registered in the language dictionary. The voice synthesizer according to claim 3, wherein the compound word is given an accent of another word having a reading and a part of speech that matches the reading and part of speech of the compound word. A speech synthesis system that has a terminal and a server.
The terminal
An input unit that acquires text data including words with ruby characters,
A communication unit that transmits the text data to the server and receives a synthetic voice signal representing the voice waveform of the text data from the server.
An audio output unit that outputs the synthesized audio signal as audio, and
Have,
The server
A communication unit that receives the text data from the terminal and transmits the synthetic voice signal to the terminal.
By performing morphological analysis on the text data, the text data is divided into combinations of morphemes.
From the morphemes included in the combination of the morphemes, the morphemes included in the phrase to which the ruby is added are extracted.
With respect to the morpheme contained in the phrase to which the ruby is added, it is determined whether or not a compound word is formed together with the morpheme continuous on the front side or the back side of the morpheme.
When the morpheme contained in the phrase to which the ruby is added forms the compound word, an accent is given to the compound word as a whole to generate a phonetic character string representing the reading of the entire text data and the position of the accent. And
A processing unit that generates the synthetic voice signal based on the phonetic character string, and
Speech synthesis system with. By performing morphological analysis on text data containing words with ruby characters, the text data is divided into combinations of morphemes.
From the morphemes included in the combination of the morphemes, the morphemes included in the phrase to which the ruby is added are extracted.
For the morpheme contained in the phrase to which the ruby is added, it is determined whether or not a compound word is formed together with the morpheme continuous on the front side or the back side of the morpheme.
When the morpheme contained in the phrase to which the ruby is added forms the compound word, the compound word as a whole is accented.
Speech synthesis method including that. By performing morphological analysis on text data containing words with ruby characters, the text data is divided into combinations of morphemes.
From the morphemes included in the combination of the morphemes, the morphemes included in the phrase to which the ruby is added are extracted.
For the morpheme contained in the phrase to which the ruby is added, it is determined whether or not a compound word is formed together with the morpheme continuous on the front side or the back side of the morpheme.
When the morpheme contained in the phrase to which the ruby is added forms the compound word, the compound word as a whole is accented.
A computer program for speech synthesis that lets a computer do things.

Images