JP2005250264A - Device, method, and program for pitch pattern generation - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a pitch pattern generating device etc., for realizing natural pitch variation. <P>SOLUTION: A tone of voice/pitch variation pattern table 233a contains tone of voice numbers for specifying respective tone of voices (1st voice to 4th voice), and deformed pitch variation patterns of standard pitch variation patterns, representing standard pitch variations of the respective tone of voices, which correspond to the standard pitch variation patterns. A tone of voice/pitch variation pattern selection part 232a selects a pitch variation pattern by taking not only a tone of voice of a syllable, but also tone of voices of syllables before and after it into consideration. A pitch pattern generation part 236 generates a pitch pattern based upon the selected pitch variation pattern and pitch specification information supplied from a text analysis part 220. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a technique for generating a pitch pattern of synthesized speech.

A speech synthesizer corresponding to Chinese is equipped with a function for outputting Chinese synthesized speech in accordance with Pinyin (Chinese reading converted into romaji).
Here, in Chinese, one kanji character corresponds to one syllable, and one syllable excludes the first consonant called “vowel” (consonant that comes at the beginning of the syllable) and the “vowel” called “vowel”. Vowels, double vowels, nasalized vowels, etc.).

  In order to obtain Chinese synthesized speech, it is necessary to input such initials and finals in Roman characters (pinyin input), but in Chinese, there may be a plurality of Chinese characters having the same pinyin. For example, a certain syllable “qi” has “period”, “odd”, “start”,..., Etc. Even if only pinyin is input, an intended conversion output candidate may not be obtained immediately.

  In order to solve such a problem, a pinyin input method with four voices is adopted that inputs a tone (tone information) called “four voices” that represents syllable inflection (change in pitch over time) in combination with pinyin. (For example, refer to Patent Document 1). This tone is basically composed of the first voice that maintains the pitch (pitch), the second voice that raises the pitch, the third voice that lowers the pitch and raises it again, and the fourth voice that lowers the pitch. (See FIG. 16). When inputting tone information, the numbers 1 to 4 of the first to fourth voices are added immediately after the corresponding pinyin. For example, “period” (= first voice), “odd” (= second voice), “start” (= third voice), and “container” (= fourth voice) are converted output candidates. Are input as “qi1”, “qi2”, “qi3”, and “qi4”, respectively. In this way, inputting tone information representing the type of tone together with pinyin provides a clue to uniquely identify the kanji and meaning corresponding to pinyin.

JP-A 61-27597

According to the above method, the pitch change for each syllable according to the input tone can be obtained, but the relationship between the tone of the syllable and the tone of the preceding and following syllables (for example, the tone of the syllable is “third voice”). Depending on the tone of the subsequent syllable, such as “second voice”, there arises a problem that the pitch change becomes unnatural.
In addition, some users may desire not only to change the pitch of the synthesized speech by specifying the type of tone, but also to freely change the pitch of the synthesized speech.

  The present invention has been made in view of the circumstances described above, and a first object thereof is to provide a pitch pattern generation technique for realizing a natural pitch change, and a second object thereof is a user. Is to provide a pitch pattern generation technique for realizing the intended pitch change.

  In order to solve the above-described problem, the pitch pattern generation device according to the present invention generates a pitch pattern representing a temporal change in the pitch of the synthesized speech corresponding to the text information based on the input text information. An acquisition means for acquiring, for each syllable, pitch designation information representing a reference pitch and tone information representing a type of tone from the written text information, a tone number, a standard pitch change pattern, and the standard pitch change pattern Storage means for associating and storing a deformed pitch change pattern obtained by transforming, and identifying the tone number from the acquired tone information of the syllable, and the tone information of the syllable preceding or following the syllable From the information, select either the standard pitch change pattern or the modified pitch change pattern corresponding to the tone number Selection means, based on the pitch designation information of syllable obtained with any pitch variation patterns selected, characterized by comprising a generating means for generating a pitch pattern of the sound section.

  According to this configuration, the tone number is specified from the acquired tone information (for example, “third voice”) of the syllable, and the tone information of the syllable preceding the syllable or the tone of the syllable following the syllable is included. From the information, either a standard pitch change pattern corresponding to the tone number (for example, a standard pitch change pattern of “third voice”) or a modified pitch change pattern obtained by modifying this standard pitch change pattern is selected (see FIG. 8 and FIG. 9). In this way, since the pitch change pattern considering not only the tone of the syllable but also the tone of the preceding and following syllables is selected, compared to the case where the pitch change pattern is selected only considering the tone of the syllable. It becomes possible to obtain a more natural pitch change.

  Further, the pitch pattern generation device according to the present invention is a device that generates a pitch pattern representing a temporal change in the pitch of the synthesized speech corresponding to the text information based on the input text information, Acquisition means for acquiring pitch designation information representing a reference pitch and tone information representing a tone type for each syllable from text information, storage means for storing a tone number and a standard pitch change pattern in association with each other, and acquisition Identifying the tone number from the tone information of the syllable extracted, extracting a standard pitch change pattern corresponding to the tone number, and using the extracted standard pitch change pattern as the tone information of the syllable preceding the syllable or the tone of the subsequent syllable Generating means for generating a deformation pitch change pattern by deforming based on information, and the generated deformation pitch change pattern Based on the pitch designation information ting acquired syllables, characterized by comprising a generating means for generating a pitch pattern of the sound section.

  Further, the pitch pattern generation device according to the present invention is a device that generates a pitch pattern representing a temporal change in the pitch of the synthesized speech corresponding to the text information based on the input text information, An acquisition means for acquiring pitch designation information representing a reference pitch for each syllable from text information, a detection means for detecting whether or not each syllable contains accent information, an accent symbol, and a pitch change pattern Storage means for associating, and for the syllable in which the accent information is detected, the accent symbol is identified from the accent information, and a selection means for selecting a pitch change pattern corresponding to the accent symbol is selected. Based on the pitch change pattern and the pitch information of the syllable from which the accent information is detected, the sound Characterized by comprising a generating means for generating a pitch pattern.

  According to such a configuration, for a syllable from which accent information is detected, an accent symbol is specified from the accent information, and a pitch change pattern corresponding to the specified accent symbol is selected (see FIGS. 11 and 12). As described above, since the pitch change pattern reflecting the contents of the accent information is selected, it is possible to obtain a pitch change that cannot be expressed by the patterned tone and a pitch change desired by the user.

  Further, the pitch pattern generation device according to the present invention is a device that generates a pitch pattern representing a temporal change in the pitch of the synthesized speech corresponding to the text information based on the input text information, From the text information, a first acquisition unit that acquires pitch designation information representing a reference pitch for each syllable, a detection unit that detects whether or not each syllable includes accent information, and the text information, Second acquisition means for acquiring tone information representing the type of tone for the syllable for which accent information has not been detected; first storage means for storing an accent symbol and a pitch change pattern in association with each other; tone number and pitch change Second storage means for storing a pattern in association with the syllable in which the accent information is detected; A first selection means for specifying the accent symbol and selecting a pitch change pattern corresponding to the accent symbol; and for the syllable from which the tone information is acquired, the tone number is specified from the tone information of the acquired syllable; Based on the second selection means for selecting a pitch change pattern corresponding to the tone number, the pitch change pattern selected by the first selection means, and the pitch information of the syllable from which the accent information is detected, the syllable A pitch pattern of the syllable is generated on the basis of the pitch change pattern selected by the second selection unit and the pitch information of the syllable from which the tone information is acquired. And second generation means.

  As described above, according to the present invention, a natural pitch change or a pitch change intended by the user can be realized.

Embodiments according to the present invention will be described below with reference to the drawings.
A. First Embodiment FIG. 1 is a diagram showing a functional configuration of a speech synthesizer 100 corresponding to Chinese according to the present embodiment. In the present embodiment, it is assumed that the device is mounted on a mobile terminal in which hardware resources are greatly restricted, such as a mobile phone, a PHS (Personal Handyphone System: registered trademark), and a PDA (Personal Digital Assistance). It can be applied to various electronic devices.

The input unit 210 supplies text information input from an operation unit (not shown) to the text analysis unit 220. 2 and 3 are diagrams illustrating text information input using the four-voice pinyin input method.
Text information is broadly classified into first type text information (see FIG. 2) and second type text information (see FIG. 3), and the pitch of synthesized speech (for example, 200 (Hz)) is designated for each text information. Pitch designation information (not shown) and the like to be included.
The first type text information is text information that does not include an accent mark, which will be described later, and is obtained by adding tone information to pinyin (hereinafter collectively referred to as “pinned information with voice tone”; see FIG. 2A). Are further added with a long sound symbol (hereinafter collectively referred to as “pinned-in information with tone and long sound”; see B in FIG. 2)).

For example, the text information “xiang1gang3 (= Hong Kong)” shown in FIG. 2A is two-syllable text information composed of pinyin information “xiang1 (= incense)” and “gang3 (= port)” with voice tone, The text information “chao1 (= super)-ren2 (= nin)” shown in FIG. 2B includes the pinyin information “chao1 (= super)-” with tone and long sound and the pinyin information “ren2 (= nin) with tone. 2 syllable text information.
Note that the long clef symbol “-” means that the syllable in which the long clef symbol exists (“chao1” in FIG. 2B) is extended by a predetermined length. The greater the number of consecutive long clef symbols, Syllable pronunciation time increases.

  On the other hand, the second type text information is text information including accent information. Accent information is information for adding inflections to the corresponding syllables. Accent symbols such as “'” and “_” and “3” and “2” indicating the strength of the inflection added after this accent symbol. ”Or the like (see FIG. 3).

  For example, the text information “'2 ye3” shown in FIG. 3A is text information of one syllable in which accent information “' 2” is added to pinyin information “ye3 (= ya)” with voice tone, and B in FIG. The text information “'3 a1 _2 −− ′ 4−” shown in FIG. 4 is accented with “' 3”, “_2”, and “'4” in the pinyin information “a1 (= A) −−−” with tone and long sound. (See FIG. 4). Note that details regarding the accent information will be described later, and a description thereof will be omitted here.

  The text analysis unit 220 analyzes the text information supplied from the input unit 210 and supplies the analysis result to the pitch generation unit 230 and the audio signal generation unit 240, respectively. More specifically, when the text analysis unit (acquisition unit, first acquisition unit) 220 receives text information from the input unit 210, the text analysis unit divides the text information for each syllable and analyzes it, thereby providing a reference for each syllable. Pitch designation information representing a pitch (for example, 200 (Hz)), phonological information representing a phoneme, and prosodic information representing a volume and length of a sound are acquired. Then, the text analysis unit 220 supplies the text information for each divided syllable to the text information type determination unit 231, supplies the acquired pitch designation information for each syllable to the pitch pattern generation unit 236, and further acquires each acquired syllable for each syllable. The phoneme information and the prosody information are supplied to the voice signal generation unit 240.

  The text information type determination unit (detection unit) 231 determines whether the text information for each phoneme supplied from the text analysis unit 220 is first type text information or second type text information. The text information type determination unit 231 determines that the text information is the first type text information when the text information does not include accent information, while the text information type determination unit 231 determines the second information when the text information includes accent information. It is determined that the text information is similar. Based on the determination result, the text information type determination unit 231 supplies the first type text information to the tone information acquisition unit 231a, and supplies the second type text information to the accent information acquisition unit 231b. As described above, in the present embodiment, if accent information is included in one syllable, the accent information is given priority regardless of whether or not tone information is included in the syllable, and processing based on the accent information is executed. However, whether priority is given to the accent information included in the syllable or the tone information can be appropriately changed according to the design of the speech synthesizer 100 or the like.

The tone information acquisition unit (acquisition unit, second acquisition unit) 231a acquires tone information for each syllable from the first type text information and supplies it to the tone / pitch change pattern generation unit 234a.
On the other hand, the accent information acquisition unit 231b acquires accent information for each syllable from the second type text information and supplies it to the accent / pitch change pattern generation unit 234b.

<Tone / Pitch Change Pattern Generation Unit 234a>
The tone / pitch change pattern generation unit 234a includes a tone / pitch change pattern selection unit (selection unit) 232a and a tone / pitch change pattern table (storage unit) 233a.
FIG. 5 is a diagram exemplifying registered contents of the tone / pitch change pattern table 233a. In the tone / pitch change pattern table (storage means, second storage means) 233a, tone numbers for specifying each tone (first to fourth voices) and pitch change patterns are registered in association with each other. Yes. The pitch change pattern represents a change in pitch over time, and a standard pitch change pattern showing a standard pitch change of each tone (see the solid line portion shown in FIGS. 8 and 9) and a corresponding standard pitch change. It is comprised by the deformation | transformation pitch change pattern which deform | transformed the pattern (refer the broken-line part shown to FIG.8 and FIG.9).

  This modified pitch change pattern is a pitch change pattern generated based on the relationship between the tone information of the preceding or succeeding syllable and the tone information of the syllable, and the modified pitch change pattern shown in FIG. 8 is other than the third voice. 9 represents the pitch change of the third voice when the syllable having the following tone is followed, and the modified pitch change pattern shown in FIG. 9 represents the pitch change of the second voice when the syllable having the first tone is preceded. (Details will be described later). In the following description, the pitch change pattern generated based on the relationship between the tone information of the preceding syllable and the tone information of the syllable is referred to as the preceding modified pitch change pattern, and the tone information of the succeeding syllable and the syllable A pitch change pattern generated on the basis of the relationship with the tone information is called a subsequent deformation pitch change pattern.

FIG. 6 is a diagram illustrating the configuration of each pitch change pattern registered in the tone / pitch change pattern table 233a.
The pitch change pattern is composed of timings t1 to tn when the time for giving a pitch change is divided into n and corresponding pitch change amounts p1 to pn. In FIG. 6, the time for applying the pitch change is divided into 101 (= n) equal parts, and the timings t1 = 0,..., T31 = 30,. .., P31 = −10,..., P101 = 30.

  FIG. 7 is a diagram illustrating a pitch change pattern obtained by linear interpolation or the like for each pitch change amount at each timing shown in FIG. As is apparent from FIGS. 6 and 7, in the present embodiment, the above timing is expressed by equally dividing the time for applying the pitch change. Can be given. In the above example, the case in which the time for giving the pitch change is equally divided is illustrated. However, the purpose is not limited to the equal division. There may be. The pitch change pattern may be fixed, but may be one that can be freely defined and changed by the user.

FIG. 8 is a diagram illustrating the pitch change pattern of the third voice, and FIG. 9 is a diagram illustrating the pitch change pattern of the second voice.
The standard pitch change pattern of the 3rd voice shows a change that once rises and then rises again (see the solid line portion shown in FIG. 8), while the subsequent modified pitch change pattern of the 3rd voice has a lower pitch. This represents a change that is maintained as it is (see the broken line portion shown in FIG. 8). By providing this third type of subsequent modified pitch change pattern, it is possible to obtain a natural pitch change even when a syllable having another tone follows the syllable of the third voice.

  When the tone / pitch change pattern selection unit (selection unit, second selection unit) 232a receives the tone information of the syllable from the tone information acquisition unit 231a, it specifies the tone number from the tone information. When the tone / pitch change pattern selection unit 232a determines that the specified tone number is “third voice”, the tone information of the subsequent syllable is referred to, and the subsequent syllable changes the tone of “third voice”. It is determined whether or not the syllable has. The tone / pitch change pattern selection unit 232a selects either the standard pitch change pattern of the third voice or the subsequent modified pitch change pattern of the third voice based on the determination result.

  For example, for the syllable “gang3 (= port)” in the syllable “wu3 (= 5)” and “xiang1 gang3 (= Hong Kong)”, the standard pitch change pattern of the third voice is selected by the tone / pitch change pattern selection unit 232a. On the other hand, the syllable “lu3 (= travel)” in “lu3 xing2 (= travel)” and the syllable “bei3 (= north)” in “bei3 jing1 (= Beijing)” The succeeding type modified pitch change pattern of the third voice is selected by the pitch change pattern selection unit 232a.

  On the other hand, the standard pitch change pattern of the second voice is a pattern that shows a change in which the pitch rises from a low position PS0 as shown in FIG. 9 (see the solid line portion shown in FIG. 9), whereas the second voice has a standard pitch change pattern. The preceding deformation pitch change pattern is a pattern representing a change in which the pitch rises from a position PS1 higher than the position PS0 (see the broken line portion shown in FIG. 9). By providing this second-type advanced pitch change pattern of the second voice, even if a syllable having the tone of the first voice, which is a high sound, precedes the normal (that is, a syllable having the tone of the first voice) It is possible to obtain a natural pitch change by starting the change from a position higher than (if not).

  It should be noted that either the preceding deformation pitch change pattern or the subsequent deformation pitch change pattern is not provided for each tone (see FIGS. 8 and 9), but the preceding deformation pitch change pattern and the subsequent deformation pitch change for each tone. Both of the patterns may be provided. In addition, the syllable referring to the tone information is not limited to the preceding or succeeding syllable as described above, but may be the syllable two or six before. Moreover, you may refer to each tone information of a plurality of syllables combining these appropriately.

  When the tone / pitch change pattern selection unit (selection unit, second selection unit) 232a receives the tone information of the syllable from the tone information acquisition unit 231a, it specifies the tone number from the tone information. When the tone / pitch change pattern selection unit 232a determines that the specified tone number is “second voice”, it refers to the tone information of the preceding syllable, and the preceding syllable changes the tone of “first voice”. It is determined whether or not the syllable has. The tone / pitch change pattern selection unit 232a selects either the standard pitch change pattern of the second voice or the preceding modified pitch change pattern of the second voice based on the determination result.

For example, the syllable “xing2 (= line)” in “lu3 xing2 (= travel)” and the syllable “rong2 (= content)” in “nei4 rong2 (= content)” have a tone / pitch change pattern. While the standard pitch change pattern of the second voice is selected by the selection unit 232a, the syllables in the syllables “quan2 (= all)” and “zhong1 wen2 (= 中文)” in “an1 quan2 (= safe)” For “wen2 (= sentence)”, the tone / pitch change pattern selection unit 232a selects the preceding modified pitch change pattern of the second voice.
The operation when the tone of the syllable is “first voice” or “fourth voice” can be described in substantially the same manner as described above, and is therefore omitted.

  When the tone / pitch change pattern selection unit 232a selects a pitch change pattern that matches the tone information from the tone / pitch change pattern table 233a, the tone / pitch change pattern selection unit 232a supplies the selected pitch change pattern to the pitch pattern generation unit 236.

<Accent / Pitch Change Pattern Generation Unit 234b>
The accent / pitch change pattern generation unit 234b includes an accent / pitch change pattern selection unit 232b and an accent / pitch change pattern table 233b.
FIG. 10 is a diagram exemplifying registered contents of the accent / pitch change pattern table 233b.
In the accent / pitch change pattern table (storage means, first storage means) 233b, accent symbols and pitch change patterns are registered in association with each other. FIG. 11 is a diagram illustrating a pitch change pattern of the accent symbol “′”, and FIG. 12 is a diagram illustrating a pitch change pattern of the accent symbol “_”.

  As shown in FIGS. 11 and 12, the pitch change pattern specified by the accent symbol “′” represents a change in which the pitch gradually increases, whereas the pitch change pattern of the accent symbol “_”. Is a pattern showing a change in which the pitch gradually decreases. These pitch change patterns may be registered in the accent / pitch change pattern table 233b, for example, as function information (for example, information representing an inclination, an intercept, etc. in the case of a straight line shown in FIG. 11). Needless to say, the pitch change pattern is not limited to a linear pattern.

  When the accent / pitch change pattern selection unit (selection unit, first selection unit) 232b receives the accent information from the accent information acquisition unit 231b, the accent / pitch change pattern selection unit 232b selects an accent symbol registered in the accent / pitch change pattern table 233b from the accent information. The pitch change pattern corresponding to the accent symbol is selected. Then, the accent / pitch change pattern selection unit 232b changes the pitch change amount indicated by the pitch change pattern (in the case of the pitch change patterns shown in FIGS. 11 and 12, the slope of the straight line) according to the accent intensity indicated by the accent information. Further, the timing is changed in accordance with the time for giving the pitch change (see below for details).

FIG. 13 is a diagram exemplifying a pitch change pattern when text information of one syllable such as “'3 a1 _2 ---' 4-” (see B in FIG. 3) is input. In FIG. 13, for convenience of explanation, a pitch change pattern is illustrated with the time for giving the pitch change being 100.
As shown in FIG. 13, the time to give the pitch change is divided into four equal parts by “a1”, “−”, “−”, “−”, and the pitch change is made by the accent information “'3” added to “a1”. ch1 is obtained, and pitch changes ch2 and ch4 are obtained by the accent information “_2” and “′ 4” added to the first and third long sound symbols “−” following this, respectively. However, since accent information is not added to the second long sound symbol “-”, the pitch change ch3 is maintained with the pitch maintained at a constant value.

The accent / pitch change pattern selection unit 232b supplies the pitch pattern generation unit 236 with the selection / change of the pitch change pattern suitable for the accent information from the accent / pitch change pattern table 233b.
The pitch pattern generation unit (generation unit, first generation unit, second generation unit) 236 includes a pitch change pattern output from the tone / pitch change pattern generation unit 234a or the accent / pitch change pattern generation unit 234b, and a text analysis unit. For example, a pitch pattern as shown in FIG. 14 is generated by adding the pitch change pattern to the reference designated pitch based on the pitch designation information extracted from the pitch change pattern supplied from 220.

  The speech signal generation unit 240 generates a synthesized speech signal based on the pitch pattern supplied from the pitch pattern generation unit 236 and the phoneme information and prosodic information supplied from the text analysis unit 220. As a result, synthesized speech based on the pitch pattern generated as described above is output to the outside via a speaker (not shown) or the like.

As described above, according to the speech synthesizer according to the present embodiment, a pitch change pattern is selected in consideration of not only the tone of the syllable but also the tone of the preceding and following syllables. As a result, it is possible to obtain a synthesized speech showing a more natural pitch change as compared with the case where the pitch change pattern is selected considering only the tone of the syllable.
When the input text information includes accent information, a pitch change pattern reflecting the accent symbol and accent strength indicated in the accent information is generated. As a result, it is possible to obtain a synthesized speech indicating a pitch change that cannot be expressed by a patterned tone or a pitch change desired by the user.

B. Modification <Modification 1>
In the above-described embodiment, the case where the tone of each syllable is classified into “four voices” having four types of characteristic pitch changes has been described. However, the tone of the syllable of Chinese (normal speech) is determined. There is also a so-called “light voice” that is lightly pronounced without changing pitch. Such a light voice is marked only by pinyin without adding tone information (eg, “xie4 xie”), and the pitch change pattern related to the preceding syllable is maintained as it is for such a light voice. Anyway. In this embodiment, Chinese is assumed, but it can be applied to any language having a tone such as Thai and Vietnamese. Moreover, although the said embodiment demonstrated the case where text information was input by pinyin, you may input text information by a Chinese character. In this case, the tone may be input using tone information as in the present embodiment, and a kanji / tone table in which each kanji is associated with the tone is prepared in advance. You may make it identify the tone of the input Chinese character by referring to a table.

<Modification 2>
FIG. 15 is a diagram illustrating a configuration of a tone / pitch change pattern generation unit 234a ′ according to the second modification. The tone / pitch change pattern generation unit 234a ′ includes a modified pitch change pattern generation unit (generation unit) 232a ′ and a tone / pitch change pattern table (storage unit) 233a ′.
In the tone / pitch change pattern table 233a ′, a tone number for identifying each tone (first to fourth voices) and a standard pitch change pattern representing a standard pitch change of each tone are registered in association with each other. However, it differs from the tone / pitch change pattern table 233a shown in FIG. 5 in that the modified pitch change patterns are not registered in association with each other.

  On the other hand, the modified pitch change pattern generation unit (generation means) 232a ′ modifies the standard pitch change pattern extracted from the tone / pitch change pattern table 233a ′, thereby changing the modified pitch change pattern (the broken line portion in FIGS. 8 and 9). Reference). Specifically, the modified pitch change pattern generation unit 232a 'first specifies a tone number based on the tone information supplied from the tone information acquisition unit 231a. Then, the modified pitch change pattern generation unit 232a 'extracts the standard pitch change pattern corresponding to the specified tone number from the tone / pitch change pattern table 233a'.

  When the modified pitch change pattern generation unit 232a ′ extracts the standard pitch change pattern, the modified pitch change pattern generation unit 232a ′ refers to the tone information of the syllable preceding the syllable (or the tone information of the subsequent syllable) and determines whether to generate the modified pitch change pattern. To decide. This determination may be made with reference to a memory or the like in which rules (deformation rules) for generating a deformation pitch change pattern are registered in advance. When the deformation pitch change pattern generation unit 232a 'determines to generate the deformation pitch change pattern, the deformation pitch change pattern generation unit 232a' refers to a deformation rule stored in a memory (not shown) or the like and appropriately deforms the standard pitch change pattern. In this manner, the modified pitch change pattern generation unit 232a ′ generates the modified pitch change pattern shown in FIGS. 8 and 9 and supplies it to the pitch pattern generation unit 236. Note that the operation after the deformation pitch change pattern generation unit 232a 'generates the deformation pitch change pattern can be described in the same manner as in this embodiment, and thus the description thereof is omitted.

<Modification 3>
Further, since the various functions related to the speech synthesizer 100 described above are realized by the CPU (or DSP) executing a program stored in a memory such as a ROM, the recording medium such as a CD-ROM is used for the program. It may be recorded and distributed on the Internet or distributed via a communication network such as the Internet.

It is a block diagram which shows the function structure of the speech synthesizer which concerns on this embodiment. It is the figure which illustrated the text information input using the pinyin input method with four voices concerning the embodiment. It is the figure which illustrated the text information input using the pinyin input method with four voices concerning the embodiment. It is the figure which illustrated the text information before and after the accent information provision which concerns on the same embodiment. It is the figure which illustrated the registration content of the tone and pitch change pattern table which concerns on the same embodiment. It is a figure which shows the structure of the pitch change pattern which concerns on the same embodiment. It is the figure which illustrated the pitch change pattern concerning the embodiment. It is the figure which illustrated the pitch change pattern of the 3rd voice concerning the embodiment. It is the figure which illustrated the pitch change pattern of the 2nd voice concerning the embodiment. It is the figure which illustrated the accent and pitch change pattern table which concerns on the same embodiment. It is the figure which illustrated the pitch change pattern of the accent symbol concerning the embodiment. It is the figure which illustrated the pitch change pattern of the accent symbol concerning the embodiment. It is the figure which illustrated the pitch change pattern of the accent symbol string which concerns on the same embodiment. It is the figure which illustrated the pitch pattern concerning the embodiment. It is the figure which illustrated the composition of the tone and pitch pattern generation part concerning modification 2. It is the figure which illustrated the pitch change pattern of each tone in Chinese.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Speech synthesizer, 210 ... Input part, 220 ... Text analysis part, 230 ... Pitch generation part, 231 ... Text information type judgment part, 231a ... Tone information acquisition part, 231b ... accent information acquisition unit, 232a ... tone / pitch change pattern selection unit, 232a '... deformed pitch change pattern generation unit, 232b ... accent / pitch change pattern selection unit, 233a, 233a' ..Voice tone / pitch change pattern table, 233b ... accent / pitch change pattern table, 234a ... Voice tone / pitch change pattern generation unit, 234b ... accent / pitch change pattern generation unit, 236 ... pitch pattern Generation unit, 240... Audio signal generation unit.

Claims (6)

An apparatus for generating a pitch pattern that represents a temporal change in pitch of synthesized speech corresponding to text information, based on input text information,
Acquisition means for acquiring, for each syllable, pitch designation information representing a reference pitch and tone information representing a type of tone from the text information;
Storage means for storing a tone number, a standard pitch change pattern, and a modified pitch change pattern obtained by deforming the standard pitch change pattern in association with each other;
The tone number is identified from the acquired tone information of the syllable, and the standard pitch change pattern or the modified pitch change corresponding to the tone number from the tone information of the syllable preceding the syllable or the tone information of the subsequent syllable A selection means for selecting one of the patterns;
A pitch pattern generation apparatus comprising: generation means for generating a pitch pattern of the syllable based on any selected pitch change pattern and the acquired pitch designation information of the syllable. An apparatus for generating a pitch pattern that represents a temporal change in pitch of synthesized speech corresponding to text information, based on input text information,
Acquisition means for acquiring, for each syllable, pitch designation information representing a reference pitch and tone information representing a type of tone from the text information;
Storage means for storing the tone number and the standard pitch change pattern in association with each other;
The tone number is identified from the acquired tone information of the syllable, a standard pitch change pattern corresponding to the tone number is extracted, and the extracted standard pitch change pattern is used as the tone information of the syllable preceding the syllable or the subsequent syllable. Generating means for generating a deformation pitch change pattern by deforming based on the tone information;
A pitch pattern generating device comprising: generating means for generating a pitch pattern of the syllable based on the generated modified pitch change pattern and the acquired pitch designation information of the syllable. An apparatus for generating a pitch pattern that represents a temporal change in pitch of synthesized speech corresponding to text information, based on input text information,
An acquisition means for acquiring pitch designation information representing a reference pitch for each syllable from the text information;
Detecting means for detecting whether or not accent information is included in each syllable;
Storage means for storing the accent symbol and the pitch change pattern in association with each other;
Selecting means for identifying the accent symbol from the accent information and selecting a pitch change pattern corresponding to the accent symbol for the syllable from which the accent information is detected;
A pitch pattern generating apparatus comprising: generating means for generating a pitch pattern of the syllable based on the selected pitch change pattern and the pitch designation information of the syllable from which the accent information is detected. An apparatus for generating a pitch pattern that represents a temporal change in pitch of synthesized speech corresponding to text information, based on input text information,
First acquisition means for acquiring, for each syllable, pitch designation information representing a reference pitch from the text information;
Detecting means for detecting whether or not accent information is included in each syllable;
Second acquisition means for acquiring tone information representing a type of tone for the syllable from which the accent information has not been detected, from the text information;
First storage means for storing an accent symbol and a pitch change pattern in association with each other;
Second storage means for storing a tone number and a pitch change pattern in association with each other;
First selection means for identifying the accent symbol from the accent information and selecting a pitch change pattern corresponding to the accent symbol for the syllable from which the accent information is detected;
For the syllable from which the tone information is acquired, the tone number is specified from the tone information of the acquired syllable, and the pitch selection pattern corresponding to the tone number is selected. The second selection unit is selected by the first selection unit. A first generation means for generating a pitch pattern of the syllable based on the pitch change pattern and the pitch designation information of the syllable from which the accent information is detected; and the pitch change pattern selected by the second selection means; A pitch pattern generation device comprising: second generation means for generating a pitch pattern of the syllable based on the pitch designation information of the syllable from which the tone information has been acquired. A method for generating a pitch pattern that represents a temporal change in pitch of synthesized speech corresponding to text information, based on input text information,
The storage means stores a tone number, a standard pitch change pattern, and a modified pitch change pattern obtained by modifying the standard pitch change pattern in association with each other.
An acquisition process for acquiring, for each syllable, pitch designation information representing a reference pitch and tone information representing a type of tone from the text information;
The tone number is identified from the acquired tone information of the syllable, and the standard pitch change pattern or the modified pitch change corresponding to the tone number from the tone information of the syllable preceding the syllable or the tone information of the subsequent syllable A selection process to select one of the patterns,
A pitch pattern generation method comprising: generating a pitch pattern of the syllable based on any selected pitch change pattern and the acquired pitch designation information of the syllable. A computer comprising storage means for storing a tone number, a standard pitch change pattern, and a modified pitch change pattern obtained by deforming the standard pitch change pattern in association with each other;
Acquisition means for acquiring, for each syllable, pitch designation information that represents a reference pitch and tone information that represents a type of tone from input text information;
The tone number is identified from the acquired tone information of the syllable, and the standard pitch change pattern or the modified pitch change corresponding to the tone number from the tone information of the syllable preceding the syllable or the tone information of the subsequent syllable A selection means for selecting one of the patterns;
A pitch pattern generation program for functioning as generation means for generating a pitch pattern of a syllable based on any selected pitch change pattern and acquired pitch designation information of the syllable.

Images