JP3421963B2 - Speech component creation method, speech component database and speech synthesis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To allow a synthesized voice to be made by a simple method and be easy to hear with a low parts count. SOLUTION: A voice of reading a sentence at a constant speed 'SHIAWASE HA ARUI-...', of which a row of the phoneme and meter of the character string is clear, is recorded in a medium 8; a time length for reproducing the first two syllables 'SHIA' is measured; at intervals of this time length (for instance 200msec), the reproduced voice on the medium 8 is successively cut to voice waveforms of each two syllables 8a, 8b,...; the character string of the recorded sentence is made to correspond to these voice waveforms 8a, 8b,..., to which a phoneme 'SHIA-' and a meter (ascending type), a phoneme 'WASE' and a meter (ascending type),... are added; and these are stored in addresses of different databases.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice component creating method for producing a voice waveform used for voice synthesis as a voice component from a continuous human voice, and a voice component database storing various voice waveforms used for voice synthesis. The present invention relates to a speech synthesis method for generating synthetic speech corresponding to a Japanese reading sentence by using this speech component database.

[0002]

2. Description of the Related Art Heretofore, as a recording / editing type voice synthesizing method in which a human voice is used as a synthesized voice unit, the following first to third methods are known. The first method is to cut out a human voice and record it,
5 of Japanese such as "i", "u", "e", "o" ...
Create all Japanese one-syllable voice parts such as 0 sounds, "Kya", "Sha", etc.
This is a method of registering this in a database or the like and combining these voice parts to synthesize a Japanese reading sentence. For example, a Japanese reading sentence is synthesized by combining voice parts for each syllable such as "ni / ho / n" and "ni", "ho", and "n".

However, when a Japanese reading sentence includes long sounds or consonants such as "Tokkyo" or "Roppongi", the speech synthesized by the first method is used. Since unnaturalness is noticeable, generally, as in the second and third methods shown below, voice parts having two or more syllables are registered in the database, and the voice parts corresponding to the Japanese reading sentence are taken out from Japan. A method of generating a synthetic speech of a word reading sentence is adopted.

In the second method, a human voice is cut out from a recorded voice to create voice parts for each syllable consisting of multiple syllables, which are registered in a database, and voice parts corresponding to Japanese reading sentences are prepared. This is a method of synthesizing Japanese reading sentences by synthesizing and connecting each phrase. The speech parts for each phrase are stored in the database with various prosody changes for the same phoneme string (reading string). For example, as a phonetic part corresponding to a phrase consisting of a combination of the noun "Osaka" and a postpositional particle, there are phoneme strings "Osaka ha", "Osaka ga", "Osaka ni", "Osaka ni", and "Osaka ni" There are various prosody changes added, and these are stored in the database.

[0005] Then, by text analysis, I want to go to / tomorrow / Osaka / go to the sentence that I want to synthesize voice. (/
Indicates a phrase break. ) Is divided into phrases, "I am", "tomorrow",
The voice parts corresponding to the phoneme sequences and the prosodic sequences of the phrases "to Osaka" and "go" are retrieved from the database, and these are connected to perform voice synthesis.

[0006] A third method is a combination of voice parts corresponding to all part-of-speech words and phrases including the adjective and verb inflectional words and phrases, which are created and cut out from recorded human voices, and are registered in a Japanese reading sentence. Is a method for synthesizing voice. Also in this case, as in the case of the second method, voice parts having various changes in prosody for the same phoneme sequence are stored in the database.

[0007] Then, the sentence to be speech-synthesized is text-analyzed and I / wa / tomorrow / osaka / to / go / go. (/ Indicates a part-of-speech delimiter.) When it is decomposed for each part-of-speech, a voice component having a phoneme sequence and a prosodic sequence corresponding to the phrase of each part-of-speech is extracted from the database and connected to the Japanese reading sentence. Is voice-synthesized.

[0008]

On the other hand, the work of cutting out a voice part from a recorded human voice is troublesome because it is necessary to select a portion having a desired phoneme and prosody while observing the voice waveform. There is a demand for a method of creating audio components that requires as few audio components as possible. However, according to the above-mentioned second conventional method, a voice component having various phoneme sequences / prosodic sequences for each phrase is required, and even if only the number of combinations of nouns and particles and the number of combinations of verbs and auxiliary verbs are considered, A huge number of voice parts must be created, and unless the limited Japanese reading sentence is synthesized by voice, it takes time and effort to create the voice parts and it is difficult to put them into practical use.

Further, in the case of the third method, the number of voice parts is smaller than that of the second method, but in this case as well, in addition to at least the number of words of all parts of speech, an adjective,
Considering the inflectional phrases for things to be utilized such as verbs, the number of voice parts far exceeding the number of words recorded in the Japanese dictionary (specifically, 100,000 words or more) is required. As with the method described above, it takes time and effort to create a voice component.

According to the present invention, the number of voice parts is small, it does not take time to create, and a voice part creating method capable of providing a voice part incorporating a human vocal rhythm and the voice part are stored. It was An object is to provide a voice component database and a method for synthesizing a voice using the voice component database.

[0011]

According to the method for creating a voice component of the present invention, a voice in which a sentence having a clear arrangement of character strings and phoneme and prosody is read at a constant speed is recorded, and a reproduced sound of the recorded voice is recorded by a human being. The number of beats (the number of syllables) based on the vocal rhythm, that is, the first recording / playback time corresponding to the number of syllables of the cycle of the vocal rhythm is set as a unit time, and the recorded voice is sequentially cut from the beginning at each unit time. Then, it is main to create a voice component by creating a recording piece and adding a corresponding character string of the sentence and its prosody to the recording piece in the order cut out into each piece. It is a characteristic.

In the voice component database of the present invention, the number of constituent syllables of various voice waveforms stored is the same,
Moreover, the lengths of the voice waveforms are the same, and the voice waveforms are stored in different addresses according to the constituent phonemes and the prosody types.
A first voice synthesis method of the present invention is the voice component database of the present invention, a recording base sentence holding means for recording voice of a base sentence other than an insertion phrase portion in a response sentence, and voice components for assembling the insertion phrase. And a character information input step of inputting a character string applicable to the insertion phrase, and a reference to the insertion phrase parts combination dictionary. A synthetic voice assembling step of selecting and connecting each voice part corresponding to the input character string from the voice part database; and a voice synthesizing step of synthesizing the connected voice part as a synthesized voice of the insertion phrase part, An output step of inserting the synthesized voice of the insertion phrase portion into the recorded base sentence voice and outputting the voice of the response sentence.

Further, according to the second voice synthesizing method of the present invention, a storage address on the voice component database of the voice component database of the present invention and each voice component for assembling each fixed phrase with respect to various fixed phrases is stored in the voice component database. And a character information input process for inputting a character string of the fixed phrase, and each voice corresponding to the input character string by referring to the fixed phrase part combination dictionary. It is characterized by comprising a synthetic voice assembling process of selecting and connecting a component from the voice component database, a voice synthesizing process of synthesizing the connected voice component as a synthetic voice, and an output process of outputting the synthetic voice. .

Furthermore, a third voice synthesizing method of the present invention describes the voice component database of the present invention and the storage address of the voice component database for assembling each phrase with respect to various phrases in the voice component database. And a character information input process for inputting an arbitrary character string and a phrase delimiter and a pause position of the arbitrary character string.
A character string synthesis voice assembling process of selecting and connecting each voice component corresponding to the input arbitrary character string from the voice component database by referring to the phrase component combination dictionary, and the connected voice component as a synthesized voice. It is characterized by having a voice synthesizing process for synthesizing and an output process for outputting the synthesized voice.

As described above, according to the present invention, a human voice is recorded at a constant speed, and a part of the human voice is cut out for each unit time based on the human vocal rhythm to produce a voice component necessary for assembling a synthetic voice. In addition, the human voice rhythm is incorporated into the voice component, the human voice rhythm can be incorporated into the synthetic voice generated from the voice component, and the voice component can be easily created.

[0016]

According to the present invention, first, a sentence containing various phonemes and prosody and having a clear arrangement of character strings is read aloud at a constant speed and recorded on a recording medium. The sound of this recording medium is played back, and the recording / playback time corresponding to the first number of beats (the number of syllables) based on the human vocal rhythm is measured while looking at the sound waveform, that is, the number of syllables is calculated from the beginning of the playback sound. The time for one cycle of the human vocal rhythm, which is the unit, is measured, and the measured recording / playback time is set as the unit time. The recorded voice is mechanically simply cut from the beginning every unit time to create a plurality of recording pieces each having a voice waveform of a certain time length.

The phoneme and prosody of the character string included in the recording piece is added to the recording piece (speech waveform) by referring to the sentence in the order of cutting. Then, the recording pieces having various phonemes and prosody are collected to form a voice component. This voice component is stored in the voice component database.
The respective voice parts stored in this database have the same number of syllables and the same length, and are stored in different addresses depending on the combination of the phoneme and the prosody type.
As shown below, an appropriate voice component is selected and connected from this voice component database according to an input character string or phrase delimiter to generate synthetic voice.

For example, when a response sentence including an insertion phrase is to be output as voice, a voice of a sentence other than the insertion phrase is pre-recorded as a recording base sentence voice, and a character string to be applied to the insertion phrase portion is input and input. Referring to the insertion phrase combination dictionary based on the character string, select the voice component to assemble the insertion phrase from the voice component database and connect it, and synthesize the connected voice component as a synthesized voice, and The synthetic voice of the insertion phrase is inserted into the recorded voice that has already been recorded, and the response sentence is output as voice.

Further, for example, when one of several fixed phrases is output by voice, the fixed storage address on the voice component database of each voice component for assembling each fixed phrase is fixed for these fixed phrases. Described in the sentence part combination dictionary, select and connect the voice parts that assemble the voice of the fixed form sentence from the voice part database by referring to the fixed form part character string input, and connect the voice parts database. The voice component is synthesized as a synthesized voice, and the synthesized voice of the standard sentence is output.

Further, for example, when synthesizing an arbitrary sentence by voice, an arbitrary character string to be synthesized, a segment break of this arbitrary character string, and character information of a pause position are input. Based on the input character information, refer to the phrase part combination dictionary, and for each phrase of an arbitrary sentence, assemble a phrase, select a voice component from the voice component database, connect it, and finally assemble all phrases, that is, an arbitrary character string. The voice parts are selected and connected, and the connected voice parts with pauses are synthesized as a synthesized voice and the synthesized voice of the arbitrary character string is output.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a flow chart showing an embodiment of a voice component making method according to the present invention. The voice component creation method of this embodiment includes a database address setting step S1 and
Constant speed recording step S2, unit time setting step S3, recording disconnection step S4, recording small piece phonetic prosody imparting step S5,
It is configured to have an address assigning step S6 and a storing step S7.

In the database address setting step S1, the address of the voice component database 1 for storing voice components as a synthetic voice generation unit is preset. Each of the voice parts to be created is composed of two syllable and fixed length voice waveforms having various phonemes and prosody necessary for voice synthesis. FIG. 2 shows a phoneme matrix of the voice component 2 having two syllables as a partial configuration of the voice component database 1. Japanese 50 syllables, Japanese syllables, voiced sounds, dakuon, semi-voiced sounds, Japanese syllables, consonants, long sounds, and silences are arranged on the row side and the column side, respectively. The second sound is arranged. In the same figure, phonemes are shown in phonetic. Not all the elements of this phonological matrix are necessary as a unit of speech synthesis,
For example, of the phonological “ga” and “ko” of “school”,
"Kou" can be used as a substitute for the "Kou" part, and the number of voice parts 2 required for synthesizing Japanese reading sentences can be several thousand.

In the database address setting step S1, "a" is assigned to each element of this phoneme matrix as "address" and "ai" is assigned.
Addresses are set in order such as "2". Furthermore, each two syllables, which is each element of the phonological matrix, has three types of prosody change, as shown in FIG. 3, namely (a) flat type, (b) rising type, and (c) falling type. The flat type is a prosody type in which there is no change in intonation between the first and second sounds, the rising type is a prosody type in which the second sound has a stronger accent than the first sound, and the falling type is a second sound. It is a prosody type with a weaker accent than the first note.

FIG. 5 shows an overview of the voice parts database 1. 4A shows a storage form of various voice parts 2 made of a phoneme matrix having a flat prosody, and B of FIG. 4 shows a storage form of various voice parts 2 made of a phoneme matrix having an ascending prosody. , C in the figure respectively shows the storage form of various voice parts 2 consisting of a phoneme matrix having a descending prosody.

As shown in the figure, in the voice component 2, branch addresses are set according to the prosody type in addition to the addresses of the phoneme matrix. If the prosody type is flat type, the branch address is number 1, the prosody type is ascending type, the branch address is 2, and if the prosody type is the descending type, branch address is 3. For example, the address 1-1 of the voice component 2 is a flat type of the phoneme "a", and the prosody of the first sound "a" and the second sound "a" has no change in intonation. Addresses 2-3 are the descending type of the phoneme "ai", and the first sound "a".
The second note "I" is a prosody with a weaker accent.

In the constant speed recording step S2, a sentence in which the phoneme and prosody of the character string are clearly arranged is recorded at a constant speed as shown in FIG. Specifically, first, the arrangement of the phoneme sequence and the prosodic sequence is clear, and when divided into two syllables, it is possible to obtain all two syllables that have all the phonemes and prosody necessary for speech synthesis. Prepare sentences. 2 like this
Separating each syllable is based on the fact that the natural biological rhythm and vocal rhythm of human beings have two beats, like marching songs and cheering songs have two beats. The voice synthesized by the voice component 2 also has a natural rhythm so that the listener can easily hear the voice naturally. That is, the basic number of beats (one beat corresponds to one syllable) is two syllables in order to incorporate human vocal rhythm.

The sentence is read by the speaker 9 at a constant speed and recorded on the recording medium 8. In this case, the speaker 9 is preferably trained to read at a constant speed. In the unit time setting step S3, the recording pieces 8a, 8b, ...
・ ・ Set the unit time for cutting out. That is, the recording medium 8 is reproduced, and the recording / reproducing time including the first two syllables (the number of beats of the vocal rhythm) is measured while watching the voice waveform, and this recording / reproducing time is set as a unit time (for example, 200 ms).
Set as.

In the recording cutting step S4, the voice waveform recorded on the recording medium 8 is simply cut in order at each of the set unit times, and the recording pieces 8a, 8b ,.・ ・ Cut out. These recording pieces 8a,
Each of 8b, ... Includes a human vocal rhythm based on two syllables. In the recording piece phonological prosody imparting step S5, phonological elements and prosody are imparted to the recording pieces 8a, 8b ,. Since the arrangement of the phoneme sequence and the prosodic sequence of the sentence recorded on the recording medium 8 is clear as described above, the phonological sequence and the prosody of each recording piece are as follows. Can be easily read by associating with.

Then, a set of recording pieces having various phonemes and prosody necessary for generating synthetic speech are prepared, and these are set as speech parts 2. The address assigning step S6 is performed by the voice component 2
Based on the phoneme and prosody information given to
The address of the voice component 2 on the voice component database 1 for storing is specified.

In the storing step S7, the voice component 2 is fitted into the address portion on the voice component database 1 shown in FIGS. 2 and 4 based on the address information of the voice component 2. Next, a specific example of the voice component creating method having the above configuration will be described with reference to FIG. Specifically, as shown in the figure, the sentence "Happiness doesn't walk. So I'm walking." (The phonological sequence in this sentence says, "There is happiness. It's "dane", and the prosody of the "shia" part is ascending
The character string information such as "rise" in the ascending type has been clarified. This sentence shows a part of the sentence to be read by the speaker 9. ) And the speaker 9 blows at a constant speed. In this way, since the speaker 9 reads at a constant speed, the length of the pause that can be put in an appropriate position in the sentence is n syllables (n: integer) (constant speed recording step S2).

Then, the recording medium 8 is recorded as "There are happiness (pause) and no (pause), so there is (pause)." The time for reproducing the recording piece 8a including the first two syllables "Shia" of the recorded sentence is measured. For example, if this recording / reproducing time is 200 ms, this 200 ms is set as the unit time (unit time setting Step S3).

Next, the reproduced sound of the recording medium 8 is sequentially cut at every unit time (200 ms) to cut out the recording pieces 8a, 8b, ... For each two syllables (recording cutting step S4). These recording pieces 8a, 8b, ...
.. are associated with the character strings of the recorded sentences, and phonemes and prosody such as phonological “shia” prosody (ascending type) and phonological “seating” prosody (ascending type) are given in order. To do.
The recording small piece to which the phoneme and the prosody are added is defined as a voice component 2 (recording small piece phonetic prosody adding step S5).

The address on the voice component database 1 is specified from the phoneme and prosody assigned to the voice component 2. For example, the voice part 2 of the first phonological “Shia” prosody (ascending type) part is specified as the address 741-2 (address assignment step S
6). The voice component 2 to which the address is assigned is fitted and stored in the address component on the voice component database 1 (storing step S7).

Next, a voice synthesizing apparatus using the voice parts created as described above will be described. As shown in FIG. 4, this voice synthesizer is configured to have a voice component database 1, a character information input unit 3, a component combination dictionary 4, a synthesized voice assembling unit 5, and a voice output unit 7. There is. The voice component database 1 stores voice components 2 having the same various phonemes and prosody as described above.

The character information input section 3 inputs character information corresponding to a voice to be synthesized. As the character information, in addition to the character string, the phrase delimiter and the pause position are input as necessary. The component combination dictionary 4 describes combinations of the voice components 2 for various input character information so that voice synthesis is performed based on the character information input from the character information input unit 3, and each voice component 2 is a voice component database. The upper address is specified.

The synthetic voice assembling unit 5 stores the voice component 2 for generating the synthetic voice based on the input character information such as the character string, the segment break, or the pause position input from the character information input unit 3. Select from and connect.
The voice synthesizing unit 6 processes the voice components 2 connected by the synthetic voice assembling unit 5 so as to form a smooth continuous voice and synthesizes the voice component 2 as a synthetic voice.

The voice output unit 7 outputs the synthesized voice. Next, a specific example of this embodiment having the above configuration will be described. (Specific Example-1) The insertion phrase part is converted to synthetic speech. Character information input by the user is used as an insertion phrase, and the fixed part of the response sentence other than this insertion phrase is recorded in advance as a recorded basic voice. Then, based on the character information input to the character information input unit 3, a synthetic voice of the insertion phrase portion is generated, and the synthetic voice is embedded in the recording base sentence voice to output a response sentence by voice.

For example, when inquiring about a telephone number,
After outputting the sentence "Please enter the prefecture or city / county name of the person you want to research.", The user answered the prefecture / city / city name. When you want to output the voice "Is it OK with ○○○○○?" As a response sentence for confirming this response when you do so, pre-record the standard part of the sentence, "Is it OK?" By the way, "○○
The part of "○○○" is an insertion phrase of a recorded sentence, and voice is synthesized based on the character information of a specific prefecture name or city / ward name input by the user.

On the other hand, FIG. 7 shows combinations of the voice parts 2 necessary for generating the synthesized voice and addresses of the respective voice parts on the voice part database for various input character information applied to the insertion phrase part. Is described in the inserted word parts combination dictionary as the parts combination dictionary 4. In the figure, both the “Kanji” character string and the “Kana” character string are accepted as keywords as input character information of the inserted word.

For example, when the user inputs a character string such as "Okayama Prefecture" or "Okayama Ken" into the character information input unit 3, the synthetic voice assembling unit 5 inserts the character string shown in FIG. 7 based on this character string. 286-
The audio component of the "oka" portion of the address 2, the audio component of the "yama" portion of the address 2482-2, and the audio component of the "ken" portion of the address 610-2 are selected and connected.

The voice synthesizing unit 6 synthesizes the voice parts connected to the inserted phrase portion as a synthesized voice of smooth continuous voice. Then, the voice output unit 7 inserts the synthesized voice of the insertion phrase portion into the recorded base voice and outputs a response sentence as voice. When voice-synthesizing place names of prefectures and municipalities in Japan as an insertion phrase as in this concrete example-1, if there are 1175 types of voice components 2 for about 4000 actual place names, voice-synthesizing is possible. Experiments have shown that this is sufficient for the production of (Specific example-2) A fixed phrase is output.

In the case of the concrete example-1, the words in the insertion phrase part are speech-synthesized, but in the concrete example-2, the fixed sentence is voice-synthesized. In this case, you can use a fixed phrase such as "Please enter the name of the prefecture in which you would like to research." Or "Thank you." As shown in Example-1. The combination of the voice parts 2 for the purpose of being put in a proper position and the addresses of the voice parts database of the respective voice parts are the fixed sentence parts combination dictionary as the parts combination dictionary 4 shown in FIG. It is described in. The fixed phrase part combination dictionary shown in the figure can accept both Japanese character strings and kana character strings of fixed phrases as keywords.

Then, the synthesized voice assembling unit 5 refers to the fixed phrase part combination dictionary based on the character information of the fixed phrase character string input by the user to the character information input unit 3 and outputs the voice corresponding to the fixed phrase. Select and connect the component 2, and in the voice synthesizer 6,
The connected voice component is synthesized into a synthesized voice, and the voice output unit 7
To output this synthesized voice. Specifically, if you want to output the fixed phrase "This is a telephone number guide." By voice, for example, enter "This is a telephone number." In the kana character string in the text information input section 3. In the synthetic speech assembling unit 5, the phoneme "kochi" selects the phonetic part 2 at the address 716-1 based on the input character string based on the fixed sentence part combination dictionary, and the phoneme "Rawa" is 2603-2. The voice component 2 of the address is selected, and the voice component 2 is selected and connected from the voice component database 1 and so on.

In this case, when the user wants to output the fixed phrase by voice, the fixed phrase is pre-numbered without inputting all the character strings of the fixed phrase in Japanese character string or Japanese kana character string as described above. It is also possible to output a synthesized voice of a fixed phrase by simply inputting the fixed phrase number into the character information input unit 3 by using the fixed phrase number as a keyword. (Specific example-3) An arbitrary sentence is output.

When synthesizing a voice of an arbitrary sentence, the character string of this arbitrary sentence, the segment break, and the pause position are input to the character information input unit 3. On the other hand, as shown in FIG. 9, the speech component 2 as the component coupling dictionary 4 describes the voice component 2 that assembles a phrase in phrase units and the addresses on the voice component database of each voice component. In the case of the phrase-parts combined dictionary shown in the figure, keywords are set so that either Japanese character strings or kana character strings can be accepted as the phrase phrases. For example, the phrase string "I" or "I"
Is a phonetic part 2 whose phoneme "wata" is at address 2901-2.
It is described in the clause part combination dictionary that the phonological “wrinkle” part is assembled from the voice part 2 at the address 780-2.

Regarding the creation of this clause parts combination dictionary,
As shown in the above-mentioned second conventional method, there are many kinds of clauses and phrases, and a huge number is recorded in the dictionary. However, as in the above-mentioned second conventional method, this enormous number of voice parts are voiced. We do not do the heavy work of cutting out while observing the waveform, and in this case, for this huge number of phrase phrases,
Since the addresses of the respective voice parts 2 for assembling phrase clauses are simply stored and stored in a predetermined storage medium, the work of creating this dictionary can be performed easily.

In the synthetic voice assembling unit 5, the character information input unit 3
Based on a plurality of phrase phrases recognized from the character string of the arbitrary sentence and the phrase delimiter input to, the phrase component combination dictionary is referred to, and the voice component 2 required for voice synthesis for each phrase is stored in the voice component database 1 Select from and connect these. For example, if you divide the arbitrary sentence "I will go to school." Into phrases, it will become "I / go to school / go." (/ Indicates the phrase break), and "I" and " Since the pause position is entered between "to school", the character strings of the phrase "I am", the phrase "to school" and the phrase "go", and the character information of the pose position are input to the character information input unit 3. .

Then, in the synthesized voice assembling unit 5, the voice component 2 for assembling the phrase is selected from the voice component database 1 and connected to each phrase by referring to the phrase component combination dictionary. For the phrase “Iwa”, select the voice component 2 of the phoneme “wata” at address 2901-2 and the voice component 2 of the phoneme “wrinkle” at address 780-2 from the voice component database 1 and select the phrase “school”. The part "to" is the phonetic part 2 of the phoneme "ga" at the address 2520-2, the phonetic part 2 of the phoneme "ko" at the address 623-1, and the phoneme "e (silence) at the address 48-1. ) ”Portion of the voice component 2 to the voice component database 1
The voice parts 2 are selected for each phrase in order such that each of the selected voice parts 2 is connected.

Further, in the synthetic voice assembling unit 5, a pause of, for example, 100 ms is given to the connected body of the voice parts 2 for assembling all the clauses according to the designated pause position. For example, in the example of the previous sentence, a pause is added because there are punctuation marks between the phoneme “wata” “wrinkle” and the phoneme “gatsu” “ko”. The voice synthesizing unit 6 synthesizes the connected body of the voice parts 2 to which the pause has been added as a smooth continuous synthetic voice, and the voice output unit 7 outputs the synthetic voice.

As described above, according to this embodiment, the recording medium 8 is reproduced to observe the first two syllable voice waveforms which are the basis of the human vocal rhythm, and the recording / reproducing time at this time is used as a unit. If it is set as time, the audio component 2 consisting of two syllables can be cut out by simply cutting the reproduction signal of the recording medium 8 every unit time. Therefore, it is much easier than the conventional work of reproducing all sound components while reproducing the sound recording medium and observing the sound waveform.

Also, the number of voice parts 2 does not exceed 100,000 as in the second and third conventional examples, and is only several thousand, so that the time and effort for creating the voice parts 2 is further increased. I can omit it. Further, according to this embodiment, since the voice component 2 consisting of two syllables incorporating the natural rhythm of human voice with two beats is used as the unit for generating the synthetic voice, the synthetic voice formed by combining the voice components 2 is also human. The natural vocal rhythm of will be adopted, and the synthesized voice will be easy to hear.

The present invention is not limited to the above-mentioned embodiment, but may have various embodiments. For example, in each of the specific examples of the above embodiments, a Japanese character string including Kanji or a Kana character string is accepted for inputting a character string, but a keyword is set so that only a Kana character string is accepted. Is also good.

Further, in this embodiment, the voice component 2 stored in the voice component database 1 is provided with three types of intonation changes, but it is also possible to have more rich intonation changes.

[0054]

As is apparent from the above description, if the voice component creating method of the present invention is used, the voice component can be simply cut out from the recording medium in which the human voice is recorded at a constant speed every unit time. Can be created, and compared with the conventional method of cutting out all the audio parts while observing the audio waveform, the creation of the audio parts is not required.

Further, since the number of voice parts is small, it does not take much time and effort. Furthermore, since the unit time cut out from the recording medium is based on the human vocal rhythm, the human vocal rhythm can be incorporated into the voice parts cut out at each unit time. The natural vocal rhythm of human beings is also incorporated into the synthesized speech, and the synthesized speech becomes easy to hear.

[Brief description of drawings]

FIG. 1 is a diagram showing a processing procedure of an embodiment of a voice component creating method according to the present invention.

FIG. 2 is a diagram showing an example of a phoneme matrix of a voice component database.

FIG. 3 is a diagram showing three types of prosody changes.

FIG. 4 is a block diagram showing an example of the functional configuration of a speech synthesis apparatus to which the speech synthesis method of the present invention is applied.

FIG. 5 is a diagram showing an example of an overall image of a voice component database according to the present invention.

FIG. 6 is an explanatory diagram showing a method of creating a voice component.

FIG. 7 is a diagram showing an example of an insertion word part combination dictionary.

FIG. 8 is a diagram showing an example of a fixed text part combination dictionary.

FIG. 9 is a diagram showing an example of a phrase part combination dictionary.

Claims (5)

(57) [Claims] 1. A voice recorded by reading a sentence in which the arrangement of character strings and phonology and prosody are clearly read at a constant speed, and the first recording of the reproduced voice of the recorded voice for the number of syllables of the period of human vocal rhythm. The playback time is set as a unit time, the recorded voice is cut in order from the beginning for each unit time to create a recording piece, and each of these recording pieces is divided into one of the recorded pieces in the order in which they are cut out. A method of creating a voice component, which comprises creating a voice component by adding a corresponding character string and its prosody. 2. A database storing various speech waveforms, wherein each speech waveform has the same number of constituent syllables and the same length, and each speech waveform has its constituent phoneme and prosody. A voice parts database, which is stored in different addresses depending on the model. 3. The voice component database according to claim 2, a recording base sentence holding unit that records a base sentence voice other than an insertion phrase portion in a response sentence, and an address on the voice component database of each voice component that assembles the insertion phrase. And a character information input step of inputting a character string to be applied to the insertion phrase, and referring to the insertion word component combination dictionary, corresponding to the input character string. A synthesized voice assembling process of selecting and connecting each voice component from the voice component database; a voice synthesizing process of synthesizing the connected voice component as a synthesized voice of the inserted phrase part; and a synthesized voice of the inserted phrase part. An output step of outputting a voice of a response sentence by fitting the voice into the recorded base voice. 4. The voice component database according to claim 2, and a fixed sentence component combination dictionary in which addresses on the voice component database of each voice component that assembles each fixed phrase are described for various fixed phrases. And a character information input step of inputting a character string of the fixed phrase, and referring to the fixed phrase part combination dictionary, selecting and connecting each voice part corresponding to the input character string from the voice part database. A voice synthesizing method, comprising: a voice synthesizing process for synthesizing; a voice synthesizing process for synthesizing the connected voice components as a synthetic voice; and an outputting process for outputting the synthetic voice. 5. The speech component database according to claim 2, and a clause component combination dictionary in which addresses on the speech component database of each speech component for assembling each clause for various clauses are described. A character information input process of inputting a character string and a phrase delimiter and a pause position of the arbitrary character string, and referring to the phrase component combination dictionary, each voice component corresponding to the input arbitrary character string is recorded from the voice component database. A voice synthesizing method comprising: a character string synthetic voice assembling process of selecting and connecting; a voice synthesizing process of synthesizing the connected voice parts as a synthetic voice; and an outputting process of outputting the synthetic voice.