JP3297221B2 - Phoneme duration control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a phoneme duration control method used in a speech rule synthesizer for converting an arbitrary text into speech.

[0002]

2. Description of the Related Art In rule speech synthesis, synthesized speech quality,
In particular, the duration of the phoneme is important as a factor influencing naturalness, and its control method has been studied conventionally.
As a first example of the conventional phonological duration control method, there is a control method which focuses on a perceptual rhythm perception point of a single syllable, and is described in the first document “Syllable arrangement rule in rule synthesis of voice using CV syllable”. "(Ishikawa, Nakajima, Imai Proceedings of the National Meeting of the Institute of Electronics, Communication and Communication Engineers, 1984 p3-19
3) is proposed.

FIG. 6 is a diagram showing a configuration of a conventional phoneme duration control method based on this method (hereinafter referred to as a first conventional example). In the first conventional example, a single syllable length determination unit 25, a rhythm perception point storage unit 10, and a consonant length determination unit 12
And a phonological boundary determining unit 14.

Hereinafter, a phoneme duration control method in the first conventional example will be described.

A single syllable length determining unit 25 determines the time length of each single syllable in the single syllable string 1, and outputs a single syllable length data string 9 in which the single syllable string information is added to the single syllable string 1. As an example of the method of determining the syllable length in the syllable length determining unit 25, a method of making all syllable lengths the same is used. The rhythm perception point storage unit 10 stores a rhythm perception point position 11 serving as a reference of auditory rhythm perception for each single syllable type. The consonant length determination unit 12 determines the duration of each consonant type by using, for example, the average length of each consonant obtained by analyzing natural speech. As shown in FIG. 7, the phoneme boundary determination unit 14
The rhythm perception point position 11 is read from the rhythm perception point storage unit 10 in accordance with the monosyllable type, and the consonant length determination unit 12
, A consonant length 13 of a single syllable is input, and the consonants are arranged so that the single syllable length coincides with the time interval of the rhythm perception point, thereby determining a phonological boundary. The time length of the vowel is determined a posteriori based on the determined phoneme boundary, and the phoneme duration time 15 is output.

As a second conventional example of the phonological duration control method, a Japanese single syllable length disclosed in Japanese Patent Application Laid-Open No. 5-281993 as a second reference is used for controlling the energy of a single syllable vowel part. Assuming that the sound is heard at the interval between the vowel centroids, which is the point at which the integrated value is half of the integral over the entire interval, the time interval between the vowel centroids tends to be a fixed time length for each consonant There is a control method using the

FIG. 8 is a diagram showing a configuration of a conventional phoneme duration control method based on this method (hereinafter referred to as a second conventional example). In the second conventional example, a vowel center-of-gravity point time length determining unit 26, a vowel center-of-gravity point length storage unit 28, a vowel center-of-gravity phoneme boundary determining unit 30, a vowel center-of-gravity phoneme boundary determining parameter storage unit 32, It consists of.

Hereinafter, a phoneme duration control method in the second conventional example will be described with reference to FIG.

The vowel center-of-gravity time storage unit 28 stores, for example, the vowel center-of-gravity time lengths 27 before and after each consonant obtained by analysis of natural speech. The vowel center-of-gravity point time length determining unit 26 reads the vowel center-of-gravity point time length 27 from the vowel center-of-gravity point time storage unit 28 according to the consonant type of each monosyllable in the input monosyllable string 1, and reads the preceding and following vowel centroids. The point-to-point time length is determined, and a vowel center-of-gravity point-to-point length data string 29 obtained by adding this to a monosyllable string is output.

For a method of determining the duration of a preceding vowel, a consonant, and a succeeding vowel between vowel centroids, see, for example, the third document "Method of setting vowel length and consonant length in rhythm rule between vowel centroids" (Kato, Hashimoto Proceedings of the Acoustical Society of Japan 1
5-13, October 1992, pp 241-242), an example of which is described. That is, the vowel center-of-gravity phonological boundary determination parameter storage unit 32 stores a linear expression for calculating a preceding vowel length, a consonant length, and a succeeding vowel length from the time length between vowel centroids given for each vowel / vowel. A vowel center-of-gravity center-phonetic boundary determining parameter 31 represented by a coefficient is stored. The phoneme duration 15 is output by obtaining the preceding vowel, consonant, and subsequent vowel length by multiplying the time length by the read-out vowel center-of-gravity phoneme boundary determination parameter 31.

[0011]

In the first conventional example, there is a tendency that a mora (a unit of a prosody corresponding to a single syllable), which is one of the most remarkable features of the Japanese phoneme duration, is uttered almost at the same time. And a fixed position of each consonant is assigned to a syllable length of a single syllable according to a rule, and a phoneme continuation length is determined. Generally, as a method of determining the syllable length, a fixed time length is given, but according to the analysis of natural speech, the length of a single syllable determined by the rhythm perception point interval is not always constant.

However, in the first conventional example, since the configuration does not include means for reproducing the variation of the monosyllable length, the output phoneme duration is different from the duration of natural speech. As a result, there is a problem that the quality of synthesized speech is degraded in the speech rule synthesis device using this.

On the other hand, in the second conventional example, the syllable length is defined by the time interval between the vowel centroids, and the syllable length is determined by the type of consonant between vowels. Therefore, although control based on consonant type is considered as a syllable length variation factor, in order to limit the syllable length variation factor to consonant types and to provide a constant syllable length regardless of other factors However, it is not possible to reproduce the variation in syllable length when placed in a different syllable environment, and the output phoneme duration is different from the duration of natural speech, as in the first conventional example. As a result, the problem that the synthesized speech quality is degraded by the speech rule synthesizing apparatus using this is not sufficiently solved.

Further, since the syllable length of each consonant is different, the duration represented by the sum of the resulting phoneme durations varies greatly depending on the type of syllable to be input, making it difficult to control the utterance speed. There was a new problem.

The present invention has been made in order to solve the above-described problem. By giving a natural phoneme duration to an input syllable string, the synthesized speech quality of a speech rule synthesizer using the same is provided. It is an object of the present invention to provide a phoneme duration control method for improving the naturalness of the sound.

[0016]

In order to achieve the above object, an invention according to claim 1 is provided in a phoneme duration control method used in a speech rule synthesizing apparatus for converting an arbitrary text into speech. Rhythm unit extraction means for extracting a rhythm unit consisting of two syllables or one syllable by judging adjacent syllables from the single syllable string, and outputting a rhythm unit string in which information of the rhythm unit type is added to the single syllable string; Rhythm unit reference length storage means for storing the basic duration time for each unit type, and the rhythm unit string as input, read out the basic duration time corresponding to the rhythm unit type from the rhythm unit reference length storage means, and input them. A rhythm unit length determining means for determining the duration of each rhythm unit, adding the continuation time to the input rhythm unit sequence, and outputting it as a rhythm unit length data sequence; Rhythm that takes a data string as input, divides the time length of a rhythm unit consisting of two syllables into a fixed ratio, determines the time length of each syllable, and outputs a single syllable length data string consisting of a single syllable type and a single syllable length A unit division unit, a rhythm perception point storage unit for storing a rhythm perception point position for each single syllable type, a consonant length determination unit for determining a duration of a consonant, and the single syllable length data string as an input, While reading the rhythm perception point position from the perception point storage means,
From the consonant length given by the consonant length determining means, determine the phoneme boundary by arranging each consonant so that the time interval of the rhythm perception point matches each monosyllable length, and determine the duration of each phoneme. And a boundary determining means.

According to a second aspect of the present invention, there is provided a phoneme duration control method used in a speech rule synthesizing apparatus for converting an arbitrary text into a speech, wherein two syllables or two syllables are determined by judging adjacent syllables from an input monosyllable string. Rhythm unit extraction means for extracting a rhythm unit composed of one syllable and outputting a rhythm unit sequence in which information of a rhythm unit type is added to a single syllable sequence;
Rhythm unit reference length storage means for storing a basic duration for each rhythm unit type, and the rhythm unit sequence as input,
The basic duration corresponding to the rhythm unit type is read from the rhythm unit reference length storage means, the duration of each input rhythm unit is determined, and given to the input rhythm unit sequence,
A rhythm unit length determining means for outputting as a rhythm unit length data string, a dividing ratio storing means for storing a dividing ratio for each type of a single syllable constituting a syllable unit consisting of two syllables, and a rhythm unit length data string as input, According to the type of the monosyllable pair, each syllable length is determined by dividing the rhythm unit according to the division ratio read from the division ratio storage means, and a monosyllable length data string including the monosyllable type and the monosyllable length is output. Rhythm unit division means for each syllable type, rhythm perception point storage means for storing a rhythm perception point position for each single syllable type, consonant length determination means for determining the duration of consonants,
The single syllable length data string is input, the rhythm perception point position is read from the rhythm perception point storage means, and the time interval of the rhythm perception point matches each single syllable length from the consonant length given by the consonant length determination means. Phoneme boundary determining means for determining a phoneme boundary by arranging each consonant in this manner and determining the duration of each phoneme.

According to a third aspect of the present invention, there is provided the phonological duration control system according to the first or second aspect, wherein the single syllable length data string to be output is determined based on accent information input separately. Singly syllable length changing means for changing the syllable length.

According to a fourth aspect of the present invention, there is provided a phoneme duration control method used in a speech rule synthesizing apparatus for converting an arbitrary text into a speech, wherein two syllables or two syllables are determined by judging adjacent syllables from an input monosyllable string. Rhythm unit extraction means for extracting a rhythm unit composed of one syllable and outputting a rhythm unit sequence in which information of a rhythm unit type is added to a single syllable sequence;
Rhythm unit reference length storage means for storing a basic duration for each rhythm unit type, and the rhythm unit sequence as input,
The basic duration corresponding to the rhythm unit type is read from the rhythm unit reference length storage means, the duration of each input rhythm unit is determined, and given to the input rhythm unit sequence,
Rhythm unit length determining means for outputting as a rhythm unit length data sequence, and inputting the rhythm unit length data sequence, dividing the time length of a rhythm unit consisting of two syllables into a fixed ratio to determine the time length of each single syllable. Rhythm unit dividing means for outputting a single syllable length data string including a single syllable type and a single syllable length, vowel length ratio storing means for storing a vowel length ratio for each phonological chain, and a single syllable length data string as inputs. Vowel length priority phonological boundary determining means for locating a syllable boundary position in a vowel section, determining a time length from a vowel consonant boundary to a syllable boundary position from a syllable length and a vowel length ratio, and outputting a duration of each phoneme; It is characterized by having.

[0020]

According to the phonological duration control method of the first aspect,
The variation in syllable length is attributed to a vocal unit longer than mora, and the rhythm unit extracting means determines a type of two adjacent syllables from the input single syllable sequence and performs two or one syllable. A rhythm unit, which is a unit of utterance composed of, is extracted, and a rhythm unit string to which a monosyllable string rhythm unit type is added is output. The rhythm unit length determining means reads the rhythm unit reference length from the rhythm unit reference length storage means according to the rhythm unit type, determines the duration time for each rhythm unit, and adds information on each rhythm unit length to the rhythm unit sequence. Outputs the rhythm unit length data string.
The rhythm unit dividing means determines the time length of each syllable by dividing the time length of the rhythm unit consisting of two syllables in the rhythm unit into a fixed ratio, and obtains a single syllable length data sequence consisting of a single syllable type and a single syllable length. Is output. The consonant length determining means determines the duration of each consonant. The phonological boundary determining means receives the single syllable length data string as input, reads a rhythm perceived point position from the rhythm perceived point storage means according to the single syllable type, determines a rhythm perceived point position of each single syllable, and sets a consonant length by the consonant length determining means. The duration is determined, the consonants are arranged so that the time interval of the rhythm perception point matches each monosyllable length, and the phoneme boundary is determined, and the duration of each phoneme is output.

According to a second aspect of the present invention, a phonological duration control method controls fluctuation due to a combination of syllables constituting a rhythm unit.
The rhythm unit division ratio is read out from the division ratio storage means for a rhythm unit composed of two syllables, the rhythm unit is divided, each single syllable length is determined, and single syllable length data composed of a single syllable type and a single syllable length is output. I do.

According to a third aspect of the present invention, the syllable duration control system incorporates the control that the accent varies the syllable length, and the single syllable length changing means controls the input single syllable length data sequence. Then, each single syllable length is changed based on separately input accent information, and the changed single syllable length data string is output.

In the phonological duration control method according to claim 4, the vowel-length-priority phonological boundary determining means receives the single syllable length data sequence or the changed single syllable length data sequence as input and calculates a single syllable length data position from the single syllable boundary position. The vowel end point and the time length from the vowel start point to the monosyllable boundary position are read from the vowel length ratio stored in the vowel length ratio storage means, obtained by calculation, the phoneme boundary is determined, and the duration of each phoneme is determined. Output the length.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of the phoneme duration control method according to the present invention will be described below with reference to the drawings. Note that the same reference numerals are given to the same elements in the conventional example or each embodiment.

Embodiment 1 FIG. 1 is a configuration diagram showing a first embodiment of a phoneme duration control method according to the present invention. This device extracts a rhythm unit consisting of two syllables or one syllable from the input single syllable sequence by judging adjacent syllables, and outputs a rhythm unit sequence obtained by adding rhythm unit type information to the single syllable sequence. A rhythm unit extraction unit 2 as extraction means, a rhythm unit reference length storage unit 4 as rhythm unit reference length storage means for storing a basic duration length for each rhythm unit type, and a rhythm unit A rhythm unit that reads the basic duration corresponding to the rhythm unit type from the reference length storage unit 4, determines the duration of each input rhythm unit, assigns it to the input rhythm unit sequence, and outputs it as a rhythm unit length data sequence A rhythm unit length determining unit 6 as a length determining means and a rhythm unit length data string are input, and the time length of a rhythm unit consisting of two syllables is divided into a fixed ratio to be used for each unit. A rhythm unit dividing unit 8 which determines a syllable time length and outputs a single syllable length data string including a single syllable type and a single syllable length, and stores a rhythm perception point position for each single syllable type. Rhythm perception point storage unit 10 as rhythm perception point storage means, and consonant length determination unit 12 as consonant length determination means for determining the duration of consonants
And a syllable length data string as input, read the rhythm perception point position from the rhythm perception point storage unit 10, and match the time interval of the rhythm perception point with each single syllable length from the consonant length given by the consonant length determination unit 12. And a phoneme boundary determining unit 14 as a phoneme boundary determining means for determining a phoneme boundary by arranging each consonant so as to determine the duration of each phoneme.

The characteristic feature of this embodiment is that the rhythm unit composed of two syllables, that is, the length of a long syllable or two single syllables is divided at a fixed ratio to determine each single syllable length. The consonants are arranged so that the time interval of the rhythm perception point in the syllable matches the single syllable length, the phonological boundary is determined, and the phonological duration is set. The ability to reproduce the important two-syllable rhythm. This allows
A speech rule synthesizing apparatus that outputs a high-quality synthesized speech having a natural rhythm can be realized.

Next, the operation in this embodiment will be described with reference to FIG.
This will be described with reference to the explanatory diagram shown in FIG. The rhythm unit extraction unit 2 determines adjacent syllables in the input single syllable string 1 and extracts a rhythm unit composed of two syllables or one syllable. As an example of the rhythm unit, for example, the three types of units shown on page 5 of the fourth document “Japanese Example Sentences and Problem Series for Foreigners 12 Pronunciation and Listening” (Toki, Murata Aratake Publishing, 1989) Used. That is, a long syllable, two consecutive single syllables other than the long syllable, and an isolated single syllable. Here, as shown in the fourth document, a long syllable is a chain of syllables that are strongly connected, and (1) a single syllable (vowel, consonant + vowel) + vowel sound; Long vowel syllable (4) Single syllable + vowel / i /. The rhythm unit extraction unit 2 determines adjacent syllables from a single syllable string, and if the syllable string is any of the above, first determines that the syllable is a long syllable. Next, syllables other than long syllables are assumed to be two syllables continuing from the beginning, and the other syllable is assumed to be an isolated single syllable. For example, as shown in FIG. 2, for an input monosyllable string “Hanasanai”, “None” is detected as a long syllable, and for the remaining “Hanasa”, a continuous monosyllable is searched from the front, and “ “Hana” becomes two syllables, and the remaining “sa” becomes an isolated monosyllable.

The rhythm unit length determination unit 6 reads the rhythm unit reference length 5 stored in the rhythm unit reference length storage unit 4 according to the type of each rhythm unit, determines the rhythm unit length, and A rhythm unit length data sequence 7 in which information on each rhythm unit length is added to the column 3 is output. The rhythm unit dividing unit 8 divides a syllable unit consisting of two syllables, that is, a long syllable and two monosyllables at a fixed ratio, determines each monosyllable length, and generates monosyllable length data consisting of a monosyllable type and a monosyllable length. Output column 9.

The consonant length determination unit 12 determines the consonant length 13 by directly using the average length of each consonant type in a natural speech sample as described in the first document, for example. The rhythm perception point storage unit 10 previously stores a rhythm perception point position which is a reference position for listening to a syllable time length specific to each consonant, as described in the above-mentioned first document. The phonological boundary determination unit 14 reads the rhythm perception point position 11 stored in the rhythm perception point storage unit 10 according to the monosyllable type, and as shown in FIG. The consonant position is determined by placing the rhythm perception point position of each monosyllable on the given syllable boundary, the vowel is given in the remaining section to which the consonant is assigned, the phonological boundary is determined, and the phonological duration 15 is output. I do.

The consonant length determination unit 12 may use, for example, the fifth document “Setting Consonant Continuation Length in Sentence Speech” (Naoki, Sakasaka Nippon Acoustics) instead of using the above average value for each consonant type directly. As proposed in the Proceedings of the Society 2-6-20 September 1990, pp. 259-260), natural speech samples are analyzed by quantification class 1 using coefficients such as consonant types, adjacent vowel types, and coefficients. The consonant length can also be determined using the multiple regression model in which is determined.

As described above, according to the present embodiment, the phoneme duration control for giving the characteristic of the time structure of a unit on an utterance longer than a single syllable, which is characteristic as a variation factor of the single syllable length in Japanese. The speech rule synthesizing apparatus using the phoneme duration control method according to the present invention can generate synthesized speech having a natural rhythmic feeling.

Embodiment 2 FIG . FIG. 3 is a configuration diagram showing a configuration of a speech rule synthesis device employing a second embodiment of the phoneme duration control method according to the present invention. The present apparatus includes a rhythm unit extraction unit 2, a rhythm unit reference length storage unit 4, a rhythm unit length determination unit 6, a rhythm perception point storage unit 10, similar to the first embodiment.
Instead of the consonant length determination unit 12 and the phonological boundary determination unit 14 and the rhythm unit division unit 8 of the first embodiment, a division ratio storage that stores a division ratio for each type of single syllable that forms a rhythm unit composed of two syllables. The rhythm unit length is determined by dividing the rhythm unit according to the division ratio read from the division ratio storage unit 16 according to the type of the single syllable pair, using the division ratio storage unit 16 as a means and the rhythm unit length data string as input. ,
A syllable-type-specific rhythm unit dividing unit 18 serving as a syllable-type-specific rhythm unit dividing unit that outputs a single-syllable-length data string including a single-syllable type and a single-syllable length.

What is characteristic in this embodiment is that each single syllable length is divided by a rhythm unit composed of two syllables, that is, a long syllable or a division length for each syllable type constituting the time length of two single syllables. The consonants are arranged so that the time interval of the rhythm perception point in each monosyllable matches the monosyllable length, the phonological boundary is determined, and the phonological duration is set. Is to be able to reproduce the rhythm of two syllables, which is important as a factor of fluctuation of the duration of the syllable. As described above, the variation of the syllable length in the rhythm unit is controlled by the division ratio for each syllable type, so that a speech rule synthesizing apparatus that outputs a high-quality synthesized speech having a natural rhythm can be realized. .

Next, the operation of this embodiment will be described.

The rhythm unit extraction unit 2 determines adjacent syllables in the input single syllable string 1, and extracts a rhythm unit composed of two consecutive single syllables other than long syllables or one syllable. Here, as the rhythm unit, for example, long syllables, two syllables, and isolated single syllables shown in the first embodiment are used. The rhythm unit length determination unit 6 reads the rhythm unit reference length 5 stored in the rhythm unit reference length storage unit 4 according to the type of each rhythm unit, determines the rhythm unit length, and A rhythm unit length data string 7 to which information of each rhythm unit length is added is output. The syllable-type-based rhythm unit division unit 18 reads the division ratio 17 stored in advance in the division ratio storage unit 16 according to the syllable type of two syllables, that is, the syllable types of long syllable and two single syllables, and divides the duration. Then, each single syllable length is determined, and a single syllable length data sequence 9 including the syllable type and each single syllable length is output.

The consonant length determining section 12 determines the consonant length 13. The phonological boundary determination unit 14 reads out the rhythm perception point position 11 stored in the rhythm perception point storage unit 10 according to the type of monosyllable, and is given by the consonant length 13 as shown in FIG. By placing the rhythm perception point position of each monosyllable on the syllable boundary, the consonant position is determined, the vowel is given in the remaining section to which the consonant is allocated, the phonological boundary is determined, and the phonological duration 15 is output.

As described above, according to the present embodiment, the phoneme duration control that gives the characteristic of the time structure of a unit on an utterance longer than a single syllable, which is characteristic as a variation factor of the single syllable length in Japanese. Syllable duration control in consideration of syllable type variation of the syllable length ratio in the rhythm unit, and speech rule synthesis using the phoneme duration control method in the present embodiment. The device can generate a synthetic voice having a natural rhythmic feeling.

Embodiment 3 FIG . FIG. 4 is a configuration diagram showing the configuration of a speech rule synthesis device employing a third embodiment of the phoneme duration control method according to the present invention. This device includes a single syllable length changing unit that changes each single syllable length based on the accent information 19 between the rhythm unit dividing unit 8 and the phonological boundary determining unit 14 in addition to the components of the first embodiment. A syllable length changing unit 20 is provided. Thereby, the monosyllable length is changed based on the accent information 19, the consonants are arranged so that the time interval of the rhythm perception point in each monosyllable matches the changed monosyllable length, and the phonological boundary is determined. Speech rule synthesis that can reproduce the rhythm of two syllables, which is important as a fluctuation factor of the duration of a single Japanese syllable, and outputs a high-quality synthetic speech with a natural rhythm because the time length is set The device can be realized.

Next, the operation of this embodiment will be described.

The rhythm unit extraction unit 2 extracts three types of rhythm units, for example, long syllables, two single syllables, and isolated single syllables from the single syllable string 1, and adds rhythm unit type information to the single syllable string 1. The rhythm unit sequence 3 is output. The rhythm unit length determination unit 6 selects the rhythm unit reference length 5 stored in the rhythm unit reference length storage unit 4 according to the type of each rhythm unit, determines each rhythm unit length, and stores A rhythm unit length data string 7 to which rhythm unit length information is added is output. The rhythm unit dividing unit 8 divides the rhythm unit length of two syllables, that is, the rhythm unit length of a long syllable or two single syllables at a fixed ratio, determines each single syllable length, and determines a single syllable length and a single syllable length. The syllable length data string 9 is output. According to the input accent information 19, the monosyllabic length changing unit 20 states, for example, "If the second monosyllable from the end of a monosyllable string has an accent nucleus, the monosyllable length at the end is reduced by 15%." Each single syllable length is changed to reflect the phenomenon seen in natural speech, and the changed single syllable length data string 21 is changed.
Is output. The consonant length determining unit 12 determines the consonant length 13. The phonological boundary determination unit 14 reads the rhythm perception point position 11 stored in the rhythm perception point storage unit 10 according to the monosyllable type, and as shown in FIG. The consonant position is determined by placing the rhythm perception point position of each monosyllable on the given syllable boundary, the vowel is given in the remaining section to which the consonant is assigned, the phonological boundary is determined, and the phonological duration 15 is output. I do.

As described above, according to the present embodiment, the phoneme duration control that gives the characteristic of the time structure of a unit on an utterance longer than a single syllable, which is characteristic as a variation factor of the single syllable length in Japanese. Can be performed, and the phoneme duration control can be performed in consideration of the syllable length variation due to the accent. In the speech rule synthesizing apparatus using the phoneme duration control method in the present embodiment, a natural rhythmic sense can be obtained. Can be generated.

Embodiment 4 FIG . FIG. 5 is a configuration diagram showing a configuration of a speech rule synthesis device employing a fourth embodiment of the phoneme duration control method according to the present invention. This apparatus includes a rhythm unit extraction unit 2, a rhythm unit reference length storage unit 4, a rhythm unit length determination unit 6, and a rhythm unit division unit 8 similar to the first embodiment.
Rhythm perception point storage unit 10 and consonant length determination unit 1 of the first embodiment
Vowel length ratio storage unit 22 as vowel length ratio storage means for storing a vowel length ratio for each phoneme chain, and a single syllable length data sequence in place of the vowel length and the vowel length. A vowel length priority phonological boundary as a vowel length priority phonological boundary determining means that is arranged in a section, determines the time length from the vowel consonant boundary to the syllable boundary position from the syllable length and the vowel length ratio, and outputs the duration of each phoneme. And a decision unit 24.

The characteristic feature of this embodiment is that the rhythm unit composed of two syllables, that is, a long syllable or two single syllables is divided at a fixed ratio to determine each single syllable length, and the single syllable length is determined. Since the boundary is in a vowel section and the duration of a syllable is set by determining the time length from a monosyllable boundary to a vowel-consonant boundary, it is important as a variation factor of the duration of a Japanese monosyllable. The syllable rhythm can be reproduced under the assumption that the barycenter position interval of the vowel is perceived as a syllable length, and a high-quality speech rule synthesizer having a natural rhythm can be realized.

Next, the operation of this embodiment will be described.

The rhythm unit extraction unit 2 extracts three types of rhythm units, for example, long syllables, two single syllables, and isolated single syllables from the single syllable string 1, and adds rhythm unit type information to the single syllable string 1. The rhythm unit sequence 3 is output. The rhythm unit length determination unit 6 selects the rhythm unit reference length 5 stored in the rhythm unit reference length storage unit 4 according to the type of each rhythm unit, determines each rhythm unit length, and stores A rhythm unit length data string 7 to which rhythm unit length information is added is output. The rhythm unit dividing unit 8 divides a rhythm unit consisting of two syllables, that is, a syllable unit length of a long syllable or two single syllables into a fixed ratio, determines each single syllable length, and comprises a single syllable type and a single syllable length. A single syllable length data string 9 is output.
The vowel priority phonological boundary determining unit 24 determines the time length from the syllable boundary to the vowel consonant boundary from the monosyllable length when the syllable boundary is arranged in the vowel section from the vowel length ratio storage unit 22 stored in advance. The vowel length ratio 23, which is a coefficient for calculation, is read, the time length from the syllable boundary to the vowel consonant boundary is determined, and the phoneme duration 15 is output.

As described above, according to the present embodiment, the phoneme duration control that gives the characteristic of the time structure of a unit on an utterance longer than a single syllable, which is characteristic as a variation factor of the single syllable length in Japanese. If the vowel ruler assigns the vowel centroid to the obtained syllable boundary, the rhythm unit control method assumes that the perception of the duration of a single syllable is the vowel centroid. In the speech rule synthesizing apparatus using the phoneme duration control method according to the present embodiment, a synthesized speech having a natural rhythm can be generated.

Embodiment 5 FIG . The rhythm unit dividing section 8 as the rhythm unit dividing means in the fourth embodiment is the same as the second embodiment.
The rhythm unit division unit 18 for each syllable type and the division ratio storage unit 16 shown in the embodiment can be replaced with each other.

Embodiment 6 FIG . The vowel-priority phoneme boundary determining unit 24 in the fourth embodiment receives the single syllable length data sequence 9 output from the rhythm unit dividing unit 8, but the single syllable length changing means in the third embodiment. Length change unit 2
0 is provided between the vowel-length-priority phoneme boundary determining unit 24 and the rhythm unit dividing unit 8, and the modified monosyllable length data string output from the monosyllable length changing unit 20 according to the accent information is changed. Instead of the syllable length data sequence, it can be input to the vowel length priority phoneme boundary determination unit 24.

Embodiment 7 FIG . The rhythm unit dividing section 8 shown in the third embodiment and the sixth embodiment can be replaced with the syllable type-based rhythm unit dividing section 18 and the dividing ratio storage section 16 in the second embodiment.

[0050]

As described above, according to the first aspect of the present invention, by determining adjacent syllables from a monosyllable string,
For example, rhythm units classified into long syllables, two single syllables, and isolated single syllables are extracted, a time length is determined for each rhythm unit type, and a rhythm unit composed of two syllables, that is, a long syllable or two single syllables is determined. Determine the length of each syllable by dividing at a fixed ratio, arrange consonants so that the time interval between rhythm perception points in each syllable matches the syllable length, determine phonological boundaries, and set the phonological duration As a result, it is possible to reproduce the rhythm of two syllables, which is important as a fluctuation factor of the duration of a single Japanese syllable.

Also, by using this, it is possible to realize a speech rule synthesizing apparatus that outputs a high quality synthesized speech having a natural rhythm.

According to the second aspect of the present invention, rhythm units classified into long syllables, two single syllables, and isolated single syllables are extracted by judging adjacent syllables from the single syllable string,
The time length is determined for each rhythm unit type, and each syllable length is determined by dividing the rhythm unit consisting of two syllables, that is, a long syllable or a division ratio for each syllable type constituting the time length of two single syllables. The consonants are arranged so that the time interval of the rhythm perception point in the syllable matches the syllable length, the phonological boundary is determined, and the phonological duration is set. As a result, it is possible to reproduce an important two-syllable rhythm.

Furthermore, since the variation of the syllable length in the rhythm unit is controlled by the division ratio for each syllable type, a speech rule synthesizing apparatus which outputs a high quality synthesized speech having a natural rhythm can be realized. It becomes possible.

According to the third aspect of the present invention, rhythm units classified into, for example, long syllables, two single syllables, and isolated single syllables are extracted by judging adjacent syllables from the monosyllable string,
The time length is determined for each rhythm unit type, the time length of a two-syllable rhythm unit, that is, a long syllable or two single syllables is divided to determine each single syllable length, and the single syllable length is changed based on accent information. The consonants were arranged so that the time interval of the rhythm perception point in each monosyllable coincided with the changed monosyllable length, the phonological boundary was determined, and the phonological duration was set. It is possible to reproduce a rhythm of two syllables which is important as a fluctuation factor of the duration.

Further, since the variation of the syllable length due to the accent can be controlled, it is possible to realize a speech rule synthesizing apparatus which outputs a high-quality synthesized speech having a natural rhythm by using this.

According to the fourth aspect of the present invention, rhythm units classified into, for example, long syllables, two single syllables, and isolated single syllables are extracted by judging adjacent syllables from the single syllable string,
The time length is determined for each rhythm unit type, and the time length of a two-syllable rhythm unit, that is, a long syllable or two single syllables is divided at a fixed ratio to determine each single syllable length. The phonological duration is set by determining the time length from the boundary of a single syllable to the boundary of a vowel consonant. It is possible to reproduce under the assumption that the barycenter position interval is perceived as a syllable length.

By using this, it is possible to realize a high-quality speech rule synthesizing apparatus having a natural rhythm.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a first embodiment of a phoneme duration control method according to the present invention.

FIG. 2 is a diagram for explaining the operation of the first embodiment.

FIG. 3 is a configuration diagram showing a second embodiment of the phoneme duration control method according to the present invention.

FIG. 4 is a configuration diagram showing a third embodiment of a phoneme duration control method according to the present invention.

FIG. 5 is a configuration diagram showing a fourth embodiment of the phoneme duration control method according to the present invention.

FIG. 6 is a configuration diagram showing a first conventional example.

FIG. 7 is an explanatory diagram showing the operation of the first conventional example.

FIG. 8 is a configuration diagram showing a second conventional example.

FIG. 9 is an explanatory diagram showing the operation of the second conventional example.

[Explanation of symbols]

1. Single syllable string, 2 rhythm unit extraction section, 3 rhythm unit string, 4 rhythm unit reference length storage section, 5 rhythm unit reference length, 6 rhythm unit length determination section, 7 rhythm unit length data string, 8 rhythm unit division section, 9 Single syllable length data string, 1
0 rhythm perception point storage unit, 11 rhythm perception point position, 1
2 consonant length determination unit, 13 consonant length, 14 phoneme boundary determination unit, 15 phoneme duration time, 16 division ratio storage unit, 1
7 division ratio, 18 syllable type rhythm unit division section, 1
9 accent information, 20 single syllable length changing section, 21 changed single syllable length data string, 22 vowel length ratio storage section, 23
Vowel length ratio, 24 Vowel length priority phonological boundary determination unit, 25
Single syllable length determining unit, 26 Vowel center of gravity time length determining unit, 2
7 vowel center-of-gravity point time length, 28 vowel center-of-gravity point length storage unit, 29 vowel center-of-gravity point length data sequence, 30 vowel center-of-gravity point phoneme boundary determining unit, 31 vowel center-of-gravity point phoneme boundary determining parameter, 32 vowel center of gravity Point to phoneme boundary determination parameter storage unit.

──────────────────────────────────────────────────続 き Continuation of front page (56) References JP-A-61-11798 (JP, A) JP-A-62-234200 (JP, A) JP-A-62-284397 (JP, A) JP-A-62-234 294295 (JP, A) JP-A-1-262599 (JP, A) JP-A-5-281993 (JP, A) JP-A-6-222793 (JP, A) JP-A-60-205497 (JP, A) (58) Field surveyed (Int.Cl. ⁷ , DB name) G10L 13/06-13/08

Claims (4)

(57) [Claims] 1. A phoneme duration control method used in a speech rule synthesizing apparatus for converting an arbitrary text into a speech, comprising: determining an adjacent syllable from an input monosyllable string;
Rhythm unit extraction means for extracting a syllable or a syllable unit consisting of one syllable and outputting a rhythm unit sequence obtained by adding rhythm unit type information to a single syllable sequence, and a rhythm unit for storing a basic duration for each rhythm unit type A reference length storage unit, the rhythm unit sequence being input, reading a basic duration corresponding to a rhythm unit type from the rhythm unit reference length storage unit, determining a duration of each input rhythm unit, A rhythm unit length determining means for assigning to a unit sequence and outputting it as a rhythm unit length data sequence; inputting the rhythm unit length data sequence as input, dividing a time length of a rhythm unit composed of two syllables into a fixed ratio, and dividing each single syllable Rhythm unit dividing means that determines the time length of each syllable and outputs a single syllable length data string consisting of a single syllable length and the position of the rhythm perception point for each single syllable type A rhythm perception point storage means for storing, a consonant length determination means for determining a duration of a consonant, a syllable length data string as an input, and reading a rhythm perception point position from the rhythm perception point storage means; The phoneme boundary is determined by arranging each consonant from the consonant length given by the length determining means so that the time interval of the rhythm perception point matches each monosyllable length, and determining the duration of each phoneme. Means, and a phoneme duration control method. 2. A phonological duration control method used in a speech rule synthesizing apparatus for converting an arbitrary text into a speech, wherein two syllable strings are determined from adjacent syllables based on an input monosyllable string.
Rhythm unit extraction means for extracting a syllable or a syllable unit consisting of one syllable and outputting a rhythm unit sequence obtained by adding rhythm unit type information to a single syllable sequence, and a rhythm unit for storing a basic duration for each rhythm unit type A reference length storage unit, the rhythm unit sequence being input, reading a basic duration corresponding to a rhythm unit type from the rhythm unit reference length storage unit, determining a duration of each input rhythm unit, Rhythm unit length determining means for giving a unit string and outputting it as a rhythm unit length data string; dividing ratio storing means for storing a dividing ratio for each type of single syllable constituting a syllable unit composed of two syllables; rhythm unit length By inputting a data string and dividing a rhythm unit according to the division ratio read from the division ratio storage means according to the type of a single syllable pair, Rhythm unit division means for each syllable type that determines each syllable length and outputs a monosyllable length data string consisting of monosyllable type and monosyllable length, and rhythm perception point storage that stores the rhythm perception point position for each single syllable type Means, a consonant length determining means for determining the duration of a consonant, and the syllable length data string as an input, a rhythm perceived point position being read from the rhythm perceived point storage means, and provided by the consonant length determining means. A phonological boundary determining means for determining a phonological boundary by arranging each consonant such that a time interval of a rhythm perception point from a consonant length matches each monosyllable length, and determining a duration of each phonological unit. A phoneme duration control method characterized by: 3. The syllable duration control system according to claim 1, wherein the output unit changes each single syllable length based on accent information separately input. A phoneme duration control method characterized by comprising changing means. 4. A phonological duration control method used in a speech rule synthesizing apparatus for converting an arbitrary text into a speech, wherein two syllables are determined from an input monosyllable string by determining adjacent syllables.
Rhythm unit extraction means for extracting a syllable or a syllable unit consisting of one syllable and outputting a rhythm unit sequence obtained by adding rhythm unit type information to a single syllable sequence, and a rhythm unit for storing a basic duration for each rhythm unit type A reference length storage unit, the rhythm unit sequence being input, reading a basic duration corresponding to a rhythm unit type from the rhythm unit reference length storage unit, determining a duration of each input rhythm unit, A rhythm unit length determining means for assigning to a unit sequence and outputting it as a rhythm unit length data sequence; inputting the rhythm unit length data sequence as input, dividing a time length of a rhythm unit composed of two syllables into a fixed ratio, and dividing each single syllable Rhythm unit dividing means for determining the time length of a syllable and outputting a single syllable length data string consisting of a single syllable type and a single syllable length, and a vowel for storing a vowel length ratio for each phonological sequence The length ratio storage means and a single syllable length data sequence are input, syllable boundary positions are arranged in vowel intervals, the time length from the vowel consonant boundary to the syllable boundary position is determined from the syllable length and the vowel length ratio, and the Vowel length priority phoneme boundary determining means for outputting a duration length.