JPH032799A - Pitch pattern coupling system for voice synthesizer - Google Patents

Abstract

PURPOSE:To improve naturalness of a synthesized voice by performing the interpolation processing based of the pitch target value and the pitch time constant of the next mora from the center point or the centroid point of the steady part of the current mora, where the pitch target value is subjected to interpolation processing by its own pitch time constant, to the next mora. CONSTITUTION:With respect to read-out of phoneme parameters, phoneme parameters are read out in the phoneme order of an input sentence by an interpolation processing part 5 after intonation and the accent type of the input sentence are determined. With respect to taking-in of the next mora, pitch data P1 to P3 and PA out of phoneme parameters of the mora following current phoneme parameters and pitch time constants DP1 to DP3 and DPA are taken in for current pitch pattern interpolation. With respect to interpolation processing of the current pitch pattern, the center point of a steady part V2 of the mora is used as the target value of current pitch data P1 to P3 and PA to perform the interpolation processing of the pitch frequency from a transition part V1 to the center point. Thus, a highly natural synthesized voice is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to a speech synthesis device using a regular synthesis method, and particularly to a method for combining pitch patterns.

B9 Summary of the Invention The present invention provides a speech synthesis device that obtains a pitch pattern through interpolation processing using a pitch target value of each phoneme and a pitch time constant. This method obtains a combination of pitch patterns that enhances the naturalness of synthesized speech.

C0 A speech synthesis device using the conventional technical rule synthesis method divides an input character string into words and phrases by syntactic analysis, determines intonation and accent for each, breaks down the words and phrases into syllables and even phonemes, and converts them into syllables. Alternatively, the sound source wave and the parameters of the articulation filter are determined for each phoneme, and synthesized speech is obtained as a response output of the articulation filter to the sound source wave.

This type of speech synthesis device has, for example, the configuration shown in FIG. The Japanese language processing unit 1 performs pronunciation conversion, etc. on the input Japanese text by referring to segment breaks and a dictionary.

The sentence processing section 2 adds intonation to sentences, and the accent processing section 3 adds accents to syllables constituting sentences and clauses.

For example, as shown in Figure 4, for the sentence input ``The cherry blossoms at school bloomed beautifully,'' the sentence intonesicon decreases from the rising point in a logarithmic manner depending on the number of syllables, and the clause accent decreases depending on the number of syllables. An accent type is determined by the phrase, and a composite intonation is obtained by synthesizing these intonations and accent types, and adding exhalation intonation, rounding by filter processing, pauses, etc.

The phoneme processing unit 4 converts each input syllable data such as rsAJ, etc. into a phoneme by referring to data in the syllable parameter storage unit 4 that defines the correspondence relationship with phonemes, which are units of vowels and consonants. For example, the syllable rSAJ is decomposed into phonemes rSJ and rAJ.

The interpolation processing unit 5 extracts phoneme parameters from the phoneme parameter storage unit 51 for each phoneme from the phoneme string data from the phoneme processing unit 4, and also extracts phoneme parameters from the phoneme parameter storage unit 51. A sound source pattern is extracted, and a sound source waveform and articulatory data are obtained from these data through interpolation processing. For example, as shown in FIG. 5, the phoneme parameters include each phoneme for consonants in three utterance time periods 01 to 0. For each time period, the duration t1~
t8, pitch P1 to P3 which is the repetition frequency of the sound source wave
, the energy of this sound source wave E l-E 3 , the sound source wave butter/G, -G, and the time constant DP of pitch and energy
Discrete data of the sound source wave is obtained using l~DP, , DE, ~DE. In addition, vowels are divided into one division OA, each with pitch P A, pitch time constant DPA, energy EA,
It has an energy time constant DEA and a sound source wave pattern GA to provide discrete data of the sound source wave. Among these, the sound source wave patterns are, for example, sound source wave patterns G, ~G as shown in FIG.
3. GA is associated with each other, and for each pattern, several tens of sample data strings are prepared in the sound source parameter storage unit 5°, and sample data of the sound source wave is extracted. In addition, the energy E I""" E j, E A defines the level of the sound source wave, that is, the loudness of the sound, and the pitches P, -P
3. PA defines the height of the frequency, that is, the pitch of the sound. The regulations for these sound source wave data are 0. ~o3,
It becomes one value in OA, and the time constant DP, ~DP,, DPASDEl~DE, , for each time period and between phonemes.
A continuous sound source wave data sequence is extracted by interpolation processing by the interpolation processing section 5 when the DEA is given.

For example, the consonant pitch P, ~P3 is given as a target value for each interval O6~03 as shown in Fig. 7, and the pitch P within each interval is indicated by a solid line or a broken line depending on the magnitude of the time constant DP, -Dp. Interpolation processing is performed n times resulting in changes as shown in .

This interpolation calculation is performed using the following recurrence formula. Go P lk+ P Ik*l
The pitch Pnk is determined as follows.

Next, the phoneme parameter storage section 5. As shown in Fig. 5, the parameters of the cross-sectional area of the sound tube model and the time constants DA, ~D
Aj and DAA are also stored. This parameter gives the parameters of the vocal tract articulation equivalent filter.
As an articulation model with seven connected articulation tubes, the cross-sectional area of each acoustic tube is A, -, ~A I? for each time period. -1% A I-t~Al
7-1, A1-3 to A17-3. These parameters are given to the articulation calculation unit 6 together with the acoustic tube time constant, and articulation calculations are performed on the sound source wave.

The articulation calculation unit 6 obtains a radiated sound waveform data string when a sound source wave is applied to an acoustic tube having a cross-sectional area parameter,
This waveform data is converted into an analog signal by the D/A converter 7, and synthesized speech is obtained from the speech output device 8.

Here, each pitch pattern found corresponding to each phoneme or syllable is combined according to the input sentence to form a pitch pattern of one phrase or word. The combination of this pitch pattern is
For example, as shown in the vowel pitch pattern shown in FIG.
The stationary part ■ is obtained by obtaining Ps and each time constant from the phoneme parameters.

The pitch pattern for each phoneme is determined by interpolation processing, and the transient part ■
3, the boundary (broken line) value between the stationary part (2) and the constant part (2) is directly dragged, and the transient part V1 is obtained by interpolating the pitch target value of the self-phoneme from the previous phoneme's transient part (3). In the figure, the <mark indicates the starting point of pitch change to the next mora.

D0 Problems to be Solved by the Invention In the conventional pitch pattern combination method, the starting point of the pitch change to the next mora becomes the boundary point of the mora, and the pitch change at the boundary point becomes a large nonlinear change. Furthermore, when there is an accent change in the mora before and after the mora boundary point, the pitch change becomes even larger. As a result, the smoothness of the synthesized speech around the mora boundaries (transient parts V, V3) is impaired, which leads to a problem of deterioration of naturalness.

SUMMARY OF THE INVENTION An object of the present invention is to provide a method for combining pitch patterns that enhances the naturalness of synthesized speech.

In order to achieve the above object, the present invention calculates a pitch target value and a pitch time constant from each phoneme parameter of an input sentence, interpolates the pitch frequency of each mora, and performs interpolation processing on the pitch frequency of each mora. In a speech synthesis device that sequentially determines each mora bitch band through processing, interpolation processing is performed using the self-pitch time constant using the pitch target value as the center point or center of gravity of the stationary section of the self-mora, and From to the next mora, interpolation processing is performed using the pitch target value and pitch time constant of the next mora, and at the boundary of each mora, the pitch target value and pitch time constant are kept the same, and the pitch frequency at the boundary part changes. smoothen.

F. Embodiment FIG. 1 is a flowchart of interpolation processing showing an embodiment of the present invention. In reading the phoneme parameters in step S1, after determining the intonation and accent type of the input sentence, the interpolation processing unit 5 reads out the phoneme parameters (FIG. 5) in the order of the phonemes of the input sentence. The acquisition of pitch data of the next mora in step S2 includes pitch data P, -P, , PA of phoneme parameters that become the next mora with respect to the parameters of the current phoneme, and pitch time constants DP, ~DP.
, , DPA is taken for self-pitch pattern interpolation. In step S3, the self-pitch pattern interpolation process converts the center point of the steady part V, of the mora into the self-pitch P, -P3.
, PA interpolation processing is performed on the pitch frequency from the transient part (3) to the center point as the target value of PA.

In the pitch pattern interpolation process in step S4, the pitch pattern from the center point of the stationary part to the transient part v3 is interpolated using the pitch time constant of the next mora as a target value. The sound source wave establishment in step S5 establishes the sound source wave from the interpolated pitch pattern and sound source wave pattern as in the conventional method.

In the pitch pattern obtained by such interpolation processing, the pitch target value is in the steady portion V, as illustrated in FIG.

The pitch due to the transient parts V, , vI becomes the frequency change due to the pitch target value of the next mora and the pitch time constant. For example, for the phoneme "o", its own pitch P is at the center point of the stationary part Y2. from the transient part vl to the steady part V,
Up to the center point, the pattern changes interpolated by the own pitch time constant DP, and from the center point of the steady part vt to the transient part V, the pitch P of the phoneme "i" becomes the next mora.
, is used as a target value and its time constant D P I is used to perform interpolation processing, and the transient portion V of the next mora "i" is performed by the same interpolation processing.

Therefore, pitch interpolation is performed with the same target value and time constant in the transient parts vl and vs, which are the joining points of the moras, resulting in a combination based on a pitch pattern in which the pitch frequency changes smoothly.

Note that when a consonant exists between phonemes, the same interpolation process is performed for the vowel, and for the consonant, an interpolated pitch pattern from the phoneme parameter is inserted following the pitch pattern of the vowel.

Further, in the embodiment, a case is shown in which the pitch target value is set as the time center of the steady portion (2), but this can also be set as the center of gravity of the mora (a position where the area is half).

Further, regarding the energy of the phoneme parameter, by similarly setting an energy target value at the center point or center of gravity of the stationary portion V and performing interpolation processing, a smooth energy change at the mora boundary point can be provided.

G8 Effects of the Invention As described above, according to the present invention, the pitch target value is set as the center or center of gravity of a stationary part, and interpolation processing is performed using the pitch time constant up to the part, and the pitch target value of the next mora is calculated from the center or center of gravity. Since interpolation processing is performed using a time constant and
The boundaries between each mora result in pitch pattern changes based on the same target value and time constant, making it possible to obtain highly natural synthesized speech due to smooth changes at the boundaries.

[Brief explanation of the drawing]

FIG. 1 is a flowchart showing one embodiment of the present invention, and FIG.
The figure shows the pitch pattern of the embodiment, Figure 3 is the configuration diagram of the speech synthesis device, Figure 4 is the intonation waveform diagram, Figure 5 is the data diagram of phoneme parameters, Figure 6 is the waveform diagram of the sound source wave pattern, and Figure 7 is the waveform diagram of the sound source wave pattern. The figure shows a pitch characteristic diagram obtained by interpolation processing, and FIG. 8 shows a conventional pitch pattern. 1...Japanese language processing section, 2...-text processing section, 3...
Accent processing unit, 4... Phoneme processing unit, 4. ... syllable parameter storage unit, 5... interpolation processing unit, 5. ...
Phoneme parameter storage unit, 5. ... Sound source parameter storage unit, 6... Articulation calculation unit, 7... D/A converter, 8.
...Audio output device. Fig. 1: Flowchart of the embodiment (2 people) Fig. 2: Pitch pattern of the embodiment: Mora Fig. 3: Speech synthesis device configuration diagram 8: Speech output device Fig. 5: Data diagram of phoneme parameters Fig. 4: Intonation waveform diagram Figure 6 Waveform diagram of sound source wave pattern Figure 7 Pitch characteristic diagram by interpolation process Figure 8 Conventional pitch pattern Relationship with case involving pitch pattern combination method correction

Claims (1)

[Claims] (1) In a speech synthesis device that calculates a pitch target value and a pitch time constant from each phoneme parameter of an input sentence, interpolates the pitch frequency of each mora, and sequentially determines each mora pitch pattern by this interpolation process, the pitch target Interpolation processing is performed using the self-pitch time constant using the value as the center point of the steady section or center of gravity of the self-mora, and interpolation processing is performed using the pitch target value and pitch time constant of the next mora from the center point of the steady section or center of gravity to the next mora. A pitch pattern combining method for a speech synthesizer characterized by performing the following steps.