JPH01200290A - Voice synthesizer - Google Patents

Abstract

PURPOSE:To attain the generation of a natural synthesized voice, which is close to a real voice, by deciding voice and silence in each vowel of an input character string on the basis of a silence generating frequency by vocal sound series and an accent pattern and a silence deciding threshold. CONSTITUTION:When the character string is inputted to a character string analyzing part 1, the analyzing part 1 determines the intonation pattern of the whole character string, the vocal sound series and accent pattern of the character string. These determined matters are outputted to a rule control part 3. On the other hand, a target feature parameter file 6 of a feature parameter file 8 outputs a target feature parameter, which expresses the feature of the vowel, to the control part 3 and a time series feature parameter file 7 outputs a time series feature parameter, which expresses the feature of a consonant, to the control part 3. The control part 3 refers the both parameters, a vocal sound control rule and a rhythm control rule from a rule file 4 and generates the parameter, which is needed for voice synthesizing, and the parameter is outputted to a voice synthesizer 5. The synthesizer 5 executes the sound synthesizing on the basis of the inputted parameter and outputs a rule synthesizing voice to correspond to the input character string.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to a speech synthesizer that generates rule-based synthesized speech.

<Prior Art> In order to generate natural synthesized speech, it is important to devoice certain vowels.

Conventionally, the rules for devoicing vowels are as follows: "Sakurai Seishin °"Things to be careful about when pronunciation of common words", Japanese Pronunciation Accent Dictionary (Revised New Edition) NHKz, Explanation/Appendix P. 12
8 1985, as the law of boat fishing.

It details the typical phonological environments in which vowel silence occurs.

In addition, the devoicing rules actually used in speech synthesizers include, for example, [Yamato Sawara 1 Kazuo Hakoda "Speech Synthesis Based on Laws" Research and Practical Application Report Vol. 27 No. 12, p. 2
562 (62) edited by Electric Corporation, 1978.

FIG. 3 is a flowchart of a devoicing determination routine using the above conventional vowel devoicing rule. This conventional devoicing determination routine will be described below with reference to FIG.

In step S3+, the target vowel is a high tongue vowel (/
i/, /u/). As a result, if it is a high-toned vowel, the process proceeds to step S32, and if not, it is determined that it is voiced and the process proceeds to step S3□.

In step S32, it is determined whether the target vowel is sandwiched between voiceless consonants. As a result, if the voice is being held, the process proceeds to step 333; if not, it is determined that there is a voice, and the process proceeds to step S37.

In step S33, the target vowel is the accent nucleus (
It is determined whether the sound has a position where the pitch relatively changes from high to low. As a result, if there is an accent kernel, the process proceeds to step S39; otherwise, the process proceeds to step S34.

In step S4, it is determined whether the target vowel is the first mora. As a result, if it is the first mora, proceed to step S31+; otherwise, proceed to step S31+.
Proceed to step 36.

In step S35, it is determined whether the preceding vowel has already been devoiced. As a result, if the voice has been devoiced, the process proceeds to step S311 as semi-devoicing; otherwise, the process proceeds to step S38.

Step S3. Then, it is determined whether the target vowel is sandwiched between voiceless fricatives of the same type. As a result, if it is interposed, the process proceeds to step 3311 as semi-devoicing, and if not, the process proceeds to step S3a.

In step S37, it is determined that the target vowel is to be voiced.

In step 5311, it is determined that the target vowel is to be devoiced.

In step S39, semi-devoicing processing such as shortening the duration of the target vowel is executed.

<Problems to be Solved by the Invention> However, the conventional devoicing rules described above do not provide clear devoicing standards between vowel devoicing and phonological sequences and between vowel devoicing and accent patterns. , the criteria for devoicing are ambiguous. Furthermore, when the rate of occurrence changes, the rate of devoicing generally changes, but the relationship between the rate of occurrence and the devoicing standard has not been quantitatively described. Therefore, the devoicing rules may not always match the actual speech. Therefore, there is a problem in that the synthesized speech generated using the conventional devoicing rules described above sounds unnatural.

Therefore, the purpose of this invention is to determine the degree of occurrence of devoicing by phoneme sequence, the degree of occurrence of devoicing by accent pattern,
Based on the devoicing threshold set according to the speaking rate,
An object of the present invention is to provide a speech synthesis device capable of generating natural synthesized speech close to real speech by determining whether a target vowel is voiced or unvoiced.

Means for Solving the Problems> In order to achieve the above object, the present invention generates speech synthesis parameters using a synthesis parameter generation means according to rules stored in a rule file from the output of a character string analysis unit into which a character string is input. In a speech synthesis device that generates a speech synthesis parameter and performs speech synthesis using a speech synthesis means based on this speech synthesis parameter,
Based on the phonetic sequence of the character string input from the character string analysis section and the degree of devoicing occurrence for each phonetic sequence stored in the rule file, the phonetic sequence devoicing occurs for each vowel in the input character string. Based on the first devoicing occurrence degree setting means for setting the degree, the accent pattern of the character string input from the character string analysis section, and the devoicing occurrence degree for each accent pattern stored in the rule file,
a second degree of devoicing occurrence setting means for setting the degree of occurrence of accent pattern devoicing for each vowel of the input character string; and a degree of occurrence of phonetic sequence devoicing set by the first degree of devoicing occurrence setting means. , a devoicing occurrence degree calculating means for calculating the devoicing occurrence degree for each vowel from the accent pattern devoicing occurrence degree set by the second devoicing occurrence degree setting means; a devoicing determination threshold setting means that stores a devoicing determination threshold for each speaking rate, and sets a devoicing determining threshold according to a specified speaking speed based on the devoicing determining threshold for each speaking speed; Based on the devoicing occurrence degree calculated by the calculating means and the devoicing judgment threshold set by the devoicing judgment threshold setting means, voicing/devoicing is determined for each vowel of the input character string. There is.

<Operation> When an arbitrary character string is input to the character string analysis section and the output from the character string analysis section is input to the synthesis parameter generation means,
Speech synthesis parameters are generated according to rules stored in a rule file.

At that time, based on the phonological sequence of the input character string and the degree of devoicing occurrence for each phonological sequence stored in the rule file, the degree of occurrence of devoicing in the phonological sequence for each vowel of the input character string is determined. is set by the first devoicing occurrence degree setting means. Also, based on the accent pattern of the input character string and the degree of devoicing occurrence for each accent pattern stored in the rule file, the degree of accent pattern devoicing for each vowel in the input character string is calculated as follows. 2. It is set by the degree of devoicing occurrence setting means. Then, based on the phoneme series devoicing occurrence degree set by the first devoicing occurrence degree setting means and the accent pattern devoicing occurrence degree set by the second devoicing occurrence degree setting means, each vowel is The devoicing occurrence degree is calculated by the devoicing occurrence degree calculating means.

Further, based on the devoicing determination threshold for each speaking speed stored in the rule file, the devoicing determining threshold at the specified speaking speed is set by the devoicing determining threshold setting means.

Then, the voicing/devoicing determining means determines the degree of devoicing of the input character based on the devoicing occurrence degree calculated by the devoicing occurrence degree calculating means and the devoicing determination threshold set by the devoicing determination threshold setting means. Determine voicing/devoicing for each vowel in the string.

Thereafter, based on the speech synthesis parameters generated by the synthesis parameter generation means and the determination result of vowel voicing/devoicing determined by the voicing/devoicing determination means,
Speech synthesis is performed by the speech synthesis means.

Therefore, it is possible to determine whether each vowel is voiced or unvoiced, taking into consideration the phoneme sequence, accent pattern, and speech rate that have a large influence on the vowel devoicing phenomenon, and it is possible to obtain natural synthesized speech.

<Embodiment> Hereinafter, the main points of the configuration and operation of the speech synthesis device of the present invention will be explained with reference to the block diagram of FIG.

When an arbitrary character string is input to the character string analysis unit 1, the character string analysis unit 1 performs syntactical interpretation of the input character string and determines the intonation pattern of the entire character string. Furthermore, the word dictionary 2 is referred to to search for words included in the character string, and the accent and phoneme sequence of each word in the character string are determined, thereby determining the phoneme sequence and accent pattern of the character string. In this way, the intonation pattern of the entire character string, the phoneme series and the accent pattern of the character string determined by the character string analysis section 1 are output to the rule control section 3.

The feature parameter file 8 is composed of a target feature parameter file 6 and a time series feature parameter file 7, and the target feature parameter file 6 includes:
Target feature parameters representing vowel features are output to the rule control section 3, and the time series feature parameter file 7 outputs time series feature parameters representing consonant features to the rule control section 3. On the other hand, the rule file 4 stores phoneme control rules for connecting the target feature parameters and time-series feature parameters output from the feature parameter file, and prosody control rules for controlling each prosody, respectively, in the rule control section. Output to 3. This prosodic control rule includes a devoiced phoneme series, a vowel devoicing occurrence degree for each phoneme series, and a vowel devoicing occurrence degree for each accent pattern, which will be described later.

The rule control unit 3 controls the phoneme control ml and each prosody for combining the target feature parameters and time series feature parameters input from the feature parameter file 8 with each phoneme input from the rule file 4. With reference to the prosodic control rules for control, the intonation pattern of the entire character string input from the character string analysis unit 1, the phonological sequence of the character string, the accent pattern, and the voicing/devoicing of vowels described below Based on the determination result, parameters necessary for speech synthesis are generated, and the generated parameters are output to the speech synthesizer 5.

The speech synthesizer 5 performs speech synthesis based on the input parameters and outputs a rule synthesis device corresponding to the input character string.

FIG. 2 is a flowchart of the voicing/devoicing determination routine carried out by the rule control section 3. The voicing/devoicing determination routine will be described below with reference to FIG.

In step S, first, the phoneme sequence is devoiced for each vowel based on the phoneme sequence manually input from the character string analysis unit 1 and the devoicing occurrence coefficient representing the degree of devoicing for each phoneme sequence stored in the rule file 4. The occurrence coefficient is determined.

However, if either the preceding or following consonant is a consonant, the above phonetic sequence devoicing occurrence coefficient is multiplied by α (α: constant of 0 < α < 1)
This is done to reduce the degree of devoicing.

Further, this phoneme sequence devoicing occurrence coefficient has, for example, the following characteristics.

1. It is a vowel or a high vowel (/i/, /u/), and the consonant preceding this vowel is a voiceless fricative (/s/, /s
h/.

/h/) or voiceless affricate (/ch/, /ls/)
and the following consonant is a voiceless plosive (/p/, /
l/, /ni/.

/py/, /ky/) or voiceless affricate (/ch/
, /ls/), the consonants before and after are voiceless plosives (
/p/, /l/.

The degree of devoicing is higher than that of /ni/, the following consonants include /py/, /ky/).

2. The vowel is a high vowel (/i/, /u/), the preceding and following consonants are voiceless plosives (/p/, /l/, /ni/, and the following consonant is /py/, /ky/). ), the preceding and following consonants are different types of unvoiced fricatives (/s/
, /sh/, /h/, the following consonant includes /hy/), the degree of devoicing is higher than that of the following consonants.

3. The vowel is a high tongue vowel (/i/, /u/), and the preceding and following consonants or different types of unvoiced fricatives (/s/, /
sh/.

/h/, the following consonant includes /by/), the preceding and following consonants are the same unvoiced fricative (/s/, /sh/,
The degree of devoicing is higher than that of /h/).

4. The vowel is a high vowel (/i/, /u/), and the consonants before and after are voiceless fricatives (/s/, /sh/).
, /h/), the degree of devoicing is higher than when either the preceding or following consonant is not a voiceless consonant.

5. Do not devoice unless the vowel is a high vowel (/i/, /u/).

Table 1 shows an example of phonological sequence devoicing occurrence coefficients divided into cases of consonants before and after a high vowel. The vertical line represents the preceding consonant, and the horizontal line represents the subsequent consonant, and the larger the number of each coefficient, the higher the degree of devoicing. Numbers in parentheses indicate cases where the consonants before and after are the same.

- Below are blank spaces - Table 1 Next, the accent pattern devoicing occurrence coefficient is calculated for each vowel from the accent pattern input from the character string analysis unit 1 and the devoicing occurrence coefficient for each accent pattern stored in the rule file 4. It will be done.

The degree of occurrence of devoicing for each accent pattern has the following characteristics.

a. Vowels in the accent nucleus are rarely devoiced.

In type b and 0 vocalizations, the degree of vowel devoicing is higher than in type l or higher vocalizations.

C0 word initials have a high degree of devoicing, except for l-type utterances.

d. Vowels immediately preceding the accent nucleus have a higher degree of devoicing than vowels in the accent nucleus, but have a lower degree of devoicing than vowels after the accent nucleus.

Table 2 shows an example of the degree of occurrence of devoicing for each accent pattern. The vertical axis indicates the accent type, the horizontal axis indicates the mora position of the target vowel, and the larger the number of each coefficient, the higher the degree of devoicing. Here, the above-mentioned accent type is classified according to the position of the syllable pronounced high among the syllables forming the word. For example, "n type" means that the second to nth syllables are pronounced high, and the first syllable is pronounced high. The syllable and all syllables below the n,100th mark indicate chanting in a low voice (Shinmeikai Japanese Encyclopedia 3rd Edition Appendix Accent).

Table 2 Furthermore, from the above-mentioned phoneme sequence devoicing occurrence coefficient and accent pattern devoicing occurrence coefficient, the devoicing occurrence degree ρ(n) (1=, l, 2...) for each vowel is determined by the following formula.

WM+l WM is the number of moras for each input phoneme sequence, and 1M+1 represents silence at the end of a word).

ρ(n) = Phonological sequence devoicing occurrence coefficient A threshold value θ, which serves as a criterion for devoicing, is determined from the speech rate input from the character string analysis unit 1 in the following manner. That is, a threshold value is determined when the speech rate is normal, and when the speech rate is faster than that, the threshold value is lowered, and when the speech rate is slower than that, the threshold value is raised.

As an example, Table 3 shows the threshold values for each speaking speed when the normal speaking speed threshold is 6.

Table 3 Also, the devoicing determination threshold θ2 when vowels that are likely to be devoiced are chained is determined by the following equation.

θ2−θ×β (β is 1<real number of β) Next, devoicing is determined for each vowel.

In step S3, n
-1.

In step S4, the devoicing occurrence degree ρ(n) of the current mora (n mora) is compared with the devoicing determination threshold θ. As a result, when ρ(n)≧θ, it is assumed that the vowel of the current mora may be devoiced, and the process proceeds to step S, where ρ(n)〈θ
In the case of , it is determined that devoicing is not to be performed and the process proceeds to step S11.

In step S, the devoicing occurrence degree ρ(n+1) of the next mora (n+1) and the devoicing occurrence degree ρ(n) of the current mora (n) are compared. As a result, ρ(n)≧ρ(n
+1) That is, if the current mora has a higher degree of devoicing, the vowel of the current mora is devoiced, but the vowel of the next mora may not be devoiced, and the process proceeds to step S8. On the other hand, if ρ(n)<ρ(n+1), that is, the degree of devoicing of the current mora is lower than that of the current mora, it is assumed that the vowel of the current mora may not be devoiced, and the process proceeds to step S6.

This is because when vowels to be devoiced follow, the vowel with a lower degree of devoicing is not devoiced to avoid unclear pronunciation.

In step S8, ρ(n+1) is increased so that the vowel of the next mora is always devoiced, and the process proceeds to step S7.

In step S7, the current mora devoicing occurrence degree ρ(n)
is compared with the devoicing determination threshold θ2 when the vowels that are likely to be devoiced are linked.

As a result, if ρ(n)≧02, the vowel of the current mora is highly devoiced, so the process proceeds to step S10.

On the other hand, if ρ(n)<02, the vowel of the current mora is not devoiced and the process proceeds to step S II.

In step S8, the next mora devoicing occurrence degree ρ(n+1
) is compared with the devoicing determination threshold θ2 when vowels that are likely to be devoiced are linked.

As a result, if ρ(n+1)<02, it is determined that the vowel of the next mora is not to be devoiced, and the process proceeds to step S.

On the other hand, if ρ(n+1)≧02, the vowel of the next mora is highly devoiced, so the process directly proceeds to step 310.

In step S, in order not to devoice the vowel of the next mora, ρ(n+1)=O is set and the process proceeds to step S10.

In step S10, it is determined that the nth mora vowel is to be devoiced, and step SI! Proceed to.

In step S II, it is determined that the n-th mora vowel is to be voiced, and the process proceeds to step Slt.

Step Sl! Then, n is incremented by one in order to move on to determining the vowel of the next mora.

In step SI3, the number of moras (
WM) or less is determined. As a result, n≦
In the case of WM, steps S4 and S12 are repeated, and if n>WM, this devoicing determination routine is ended.

As described above, in this invention, the devoicing determination threshold, which is set according to the degree of devoicing occurrence for each phoneme series, the degree of devoicing occurrence for each accent pattern, and the speaking speed, and the vowel that causes devoicing are linked. Since it determines whether the target vowel is voiced or unvoiced based on the devoicing determination threshold of can be muted. Therefore, according to the present invention, more natural synthesized speech can be generated.

<Effects of the Invention> As is clear from the above, the speech synthesis device of the present invention has the following effects:
The degree of devoicing occurrence for each vowel in the input character string is calculated based on the degree of devoicing occurrence for each phoneme series and the degree of devoicing occurrence for each accent pattern in the rule file, and further,
Based on the devoicing determination threshold for each speaking speed, set the devoicing determination threshold at the specified speaking speed, and determine whether each vowel in the input character string is voiced based on the degree of devoicing occurrence and the devoicing determination threshold. Since the method determines whether the target vowel is voiced or devoiced, it is possible to determine whether the target vowel is voiced or devoiced, taking into account the phonological sequence, accent pattern, and speech rate, which have a large influence on the vowel devoicing phenomenon. can. Therefore, it is possible to generate natural synthesized speech that is close to real speech.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the speech synthesis device of the present invention, FIG. 2 is a flowchart of a vowel voicing/devoicing determination routine in the above embodiment, and FIG. 3 is a conventional voicing/devoicing determination routine. It is a flowchart of a determination routine. l... Character string analysis unit, 2... Word dictionary, 3... Rule control unit, 4... Rule file, 5... Speech synthesizer, 6... Target feature parameter file, 7.・・・
Time series feature parameter file, 8... feature parameter file.

Claims (1)

[Claims] (1) From the output of the character string analysis unit into which a character string is input, the synthesis parameter generation means generates speech synthesis parameters according to the rules stored in the rule file, and the speech synthesis means synthesizes speech based on the speech synthesis parameters. In a speech synthesis device that performs a first devoicing occurrence degree setting means for setting the degree of devoicing occurrence in the phoneme series for each vowel; an accent pattern of the character string input from the character string analysis section; and devoicing by accent pattern stored in the rule file. a second devoicing occurrence degree setting means for setting an accent pattern devoicing occurrence degree for each vowel of the input character string based on the devoicing occurrence degree; and a devoicing occurrence degree setting means set by the first devoicing occurrence degree setting means. Devoicing occurrence degree calculating means for calculating the devoicing occurrence degree for each vowel from the phoneme series devoicing occurrence degree and the accent pattern devoicing occurrence degree set by the second devoicing occurrence degree setting means; The above rule file stores devoicing determination thresholds for each speaking rate, and based on these devoicing determining thresholds for each speaking rate, a devoicing determination threshold setting means sets a devoicing determining threshold according to a specified speaking rate. Based on the devoicing occurrence degree calculated by the devoicing occurrence degree calculation means and the devoicing determination threshold set by the devoicing determination threshold setting means, each vowel of the input character string is voiced or unvoiced. A speech synthesis device characterized by determining whether