JP2990691B2 - Speech synthesizer - Google Patents

Abstract

PURPOSE:To synthesize a voice with good quality without increasing the number of data by dividing the sectional area of an acoustic tube which simulates the voice tube of a human roughly at the part corresponding to the inner part of the throat, and controlling the part corresponding to the lips closely. CONSTITUTION:Acoustic tubes S1-S4 corresponding to the inner part of the throat are regarded as one acoustic tube of 4cm pitch, whose intervals are controlled with the same data. Similarly, the pitch is decreased stepwise so that S5-S7 are 3cm, S8-S9 are 2cm, and the subsequent part is 1cm; and the parts of acoustic tubes S12-S17 corresponding to the mouth and lips are formed by dividing finely acoustic tubes corresponding to the parts where the sectional area variation of the voice tube is large like 1cm, 0.5cm or 0.2cm. Then a delay block ZD is normalized at 1cm pitch and interposed between the equivalent circuits of two adjacent acoustic tubes. Consequently, even when the number of divisions of a variable sectional area acoustic tube close to the throat is decreased, a voice which is close to a natural word is synthesized and when the quantity which is decreased nearby the throat is increased on the lip side, the voice which is close to the natural word is obtained without increasing a memory.

Description

The present invention relates to a rule speech synthesizer for synthesizing human speech.

B. Summary of the Invention The present invention regards the vocal tract from the human vocal cords to the lips as a plurality of divided acoustic tubes of variable cross-section, and uses two acoustic tubes adjacent to the divided acoustic tubes as a propagation current source and an acoustic tube. Along with treating as a parallel circuit of surge impedance inversely proportional to the tube cross-sectional area, a delay circuit is provided for the propagation of the propagation current, and the length of the acoustic tube having the same area is changed by changing the constant of the delay circuit. In addition to this, the constant of the delay circuit is normalized to a unit that is easy to calculate and is blocked, and this delay block is inserted more frequently on the throat side where the cross-sectional area changes little when voice is generated and less on the lip side where the change is large. In this way, a voice close to a natural language can be obtained without increasing the number of memories required for calculating the current value of each part flowing through the individual equivalent circuits of the divided acoustic tubes.

C. Prior Art Electronic devices that artificially synthesize and output speech have recently become low-priced with one to several chips of speech recognition and speech synthesis LSIs using speech information processing and large-scale integrated circuit technology of semiconductors. And various methods have been proposed depending on the purpose of use and restrictions. In this speech synthesis,
Recording and editing the raw voices generated by humans, combining them appropriately and editing them into sentences, and extracting only the parameters of human voices without directly using human voices In the speech synthesis process, there is a method of artificially creating a speech signal by controlling its parameters.

There is a per call (PARCOR) method which is widely used because a high quality synthesized sound can be obtained by this parameter method.

When audio is handled by an electronic computer, the audio waveform is sampled at certain intervals, the value of the audio signal at each sampling point is converted from analog to digital, and the value is displayed with 0 and 1 codes. To record faithfully with an analog signal, it is necessary to increase the number of bits, but a speech synthesis signal requires a very large amount of memory.

Therefore, various highly efficient coding methods have been researched and developed in order to minimize the amount of information.

As one of the methods, there is a delta modulation method in which a minimum of one bit is used for information of one audio signal. This method uses one bit to determine whether the next audio signal value is higher or lower than the current value. If it is higher, a code "1" is given, and if it is lower, a code "0" is given and the audio signal is encoded. In the actual system configuration, a fixed amplitude step amount (delta) is determined, and the value of the voice obtained by the conventional coding and the input voice are set so that errors are not accumulated. Encoding is performed on the residual signal from the signal.

Such a configuration is called prediction coding, and uses a linear autocorrelation function (a prediction is made from several previous sample values) and a Percoll method (a partial autocorrelation function called a Percoll coefficient k instead of the prediction coefficient of the linear prediction method). ).

D. Problems to be Solved by the Invention As described above, a technique using predictive coding has a problem that articulation coupling corresponding to a joint between sounds is difficult. For example, in a vowel from a vowel to a vowel through a consonant, the sound of the joint between vowels and vowels is cut off in the process from the steady state of a vowel to a consonant through a transition, and from the transition of a vowel to a steady vowel sound. Does not give a natural feeling when

In the case of musical instrument sound synthesis, the seams between scales are important, but since the synthesis method differs from the principle of sound generation of actual musical instruments,
After all, there is no natural feeling, especially in reverberation. In order to approach a natural sound in both cases, there are problems such as a large number of electronic components such as a memory and an arithmetic unit constituting the natural sound, which makes the apparatus expensive.

E. Means and Action for Solving the Problems Therefore, the inventor of the present application has proposed that the generation of human sounds and the musical sounds of musical instruments are produced by changes in the shape of the human mouth and the length and cross-sectional area of the acoustic tube. Analyzing the traveling wave phenomenon representing the transmission of sound waves in the acoustic tube with an acoustic tube equivalent circuit, focusing on the fact that the cross-sectional area of the acoustic tube is inversely proportional to the surge impedance, and simulating the cross-sectional area by changing the surge impedance Speech synthesis method that enables smooth articulation coupling by continuously changing the surge impedance to facilitate the synthesis of sounds similar to human utterances and improve the naturalness of speech And filed a patent application earlier (Japanese Patent Application No. 62-91705, hereinafter referred to as the prior application).

The present invention has been made based on the invention of the prior application. In particular, as a result of analyzing the cross-sectional area change due to the movement of the human vocal tract and lips at the time of voice generation, the depth of the vocal tract (throat (Back) did not change much, and it turned out that the tip of the mouth and lips changed greatly. Focusing on this result, the present invention roughly controls the cross-sectional area of the acoustic tube simulating the human vocal tract by dividing the part corresponding to the back of the throat and making the part corresponding to the lips dense, It is an object of the present invention to provide an apparatus for synthesizing high-quality sound without increasing the number.

F. Embodiment First, the outline of the invention of the prior application which forms the basis of the present application will be described.

The change in the cross-sectional area of the vocal tract when a voice is generated is, for example, in the case of the utterance of `` A '', the vocal cords intermittently open and close the expiration with the exhalation pushed out from the lungs with the back of the throat narrow and the lips open A sound wave that is repeatedly reflected in the road (acoustic tube) comes out as a sound waveform of “A”. If "i" has a wider throat and a narrower lip, the voice waveform of "a" is output.

In this manner, the frequency is determined by the appearance of the mouth, and if the appearance of the mouth is simulated, "A" or "I" is uttered. The appearance of the mouth can be simulated by the cross-sectional area of the sound tube, and the change in the cross-sectional area of the sound tube can be simulated by the change in surge admittance. Therefore, the appearance of the mouth can be simulated by changing the surge admittance. Since the change in surge admittance can be changed very easily on an electric circuit, various voices can be synthesized by electric signals. Fig. 2 (a) shows the cross-sectional areas A ₁ and A ₂
... is obtained by simulating the vocal tract by connecting the sound tube S ₁ to S _n which have different cross-sectional areas and A _n. FIG. 3A shows a case where the acoustic impedance is replaced by an LC circuit of an electric circuit. Each acoustic tube is formed as one LC line, and the whole is formed as an n-1 electric circuit of a concentrated line. FIG. 2 (c) shows a traveling wave equivalent model diagram. In FIG. 2A, T _{1 to} T _n are:
Shows a circuit element consisting of surge impedance component of each acoustic tube S ₁ to S _n. The present invention basically relates to these second
Figure (b), by applying the circuit (c), this Yotsute to changing the impulse current and surge impedance of each circuit element T ₁ through T _n supplies, the acoustic tube model tone wave and the acoustic to correspond to changing the sectional area of the tube, Yotsute the current output from the circuit element T _n in the final stage to be supplied to the utterance section such as a speaker, a voice obtained from the acoustic tube model simulatively It is something to create.

Specifically, as shown in FIG. 2 (c), a circuit equivalent to the above-described electric circuit is assumed, and the current of the current source in this equivalent circuit is changed with time, and the calculation is performed as described later. Since the cross-sectional area ratio of the acoustic tube is introduced into the equation,
Each cross-sectional area A ₁ to A _n is varied with respect to time, it is determined by calculating the current value of the by connexion each part thereto. P In the drawing current source, Z ₀ is the impedance of the current source, Z ₁ to Z _n surge impedance of each circuit element T ₁ ~T _n, Z _L is the radiation _{impedance, i 0A ~i (n-1} ) A , i _{1B to} i _nB , a _{0A to} a _{(n-1) A} , a _1B
Aa _nB is the current of the current path corresponding to the symbol, W _0A WW
_{(n-1) A, W} 1B ~W nB current _{source, I 0A ~I (n-1} ) A is backward wave current, I _1B ~I _nB indicates the forward wave current. In this equivalent circuit, for example, when attention is paid to the binding portion of the circuit element T ₁ through T _n,
A current source W _1B corresponding to a current I _1B flowing through the circuit element T ₁ toward T ₂ and a current I _1A flowing through the circuit element T ₂ toward T ₁
Assuming a current source W _1A made to correspond to the current I _1B are circuit elements
At the boundary between T ₁ and T _2, the reflected wave current i _1B reflected to T ₁ and the transmitted wave current a _1A transmitted to T ₂ are separated, and the current I _1A is divided between the circuit elements T ₂ and T ₁ . Reflected currents i _1A and T ₁ reflected to T ₂ at
This is equivalently expressed as being divided into a transmitted wave current _a1B transmitted to the light source.

 Next, the calculation will be described.

First, the block including the first-stage current source P in FIG. 2 (c) can be considered as a superposition of two circuits as shown in FIG. 2 (d). Accordingly, when the voltage of the current source P is set to V, the currents a ₁ and a ₂ in the figure are expressed by the equations (1) and (2), respectively. As a result, the current a _0A is expressed by the equation (3). .

_{a 1 = V / Z 0 +} Z 1 ...... (1) a 2 = Z 0 / Z 0 + Z 1 · I oA ...... (2) a 0A = a 1 + a 2 = 1 / Z 0 + Z 1 (V + Z 0 · I _0A ) (3) If a current is supplied to the equivalent circuit for the first time, a _0A is obtained by setting I _0A to zero. Then, the calculation is sequentially performed based on this value. Taking the arithmetic expressions of the current values of the first block and the second block located at the left end in the drawing as examples, the current values are represented by the following expressions (4) to (12).

_{a 0 'A = 1 / Z} 0 + Z 1 (V' + Z 0 · I 0A) ...... (4) i 0 'A = a 0' A -I 0A ...... (5) I 0 'A = i 1' _{B + a 1 'B ...... (} 6) a 1' B = S 1B (I 1B + I 1A) ...... (7) i 1 'B = a 1' B -I 1B ...... (8) I 1 'B = _{i 0 'A + a 0'} A ...... (9) a 1 'A = S 1A (I 1B + I 1A) ...... (10) i 1' A = a 1 'A -I 1B ...... (11) I 1 If _{'a = i 2' B +} a 2 'B ...... (12) to advance such a calculation, the calculation equation for block of the last stage may be expressed as (13) to (15).

_{a n 'B = Z L /} Z n + Z L · I nB ...... (13) i n' B = a n 'B -I nB I n' B = i (n-1) A = a (n-1 _{) a} ...... ₍₁₄₎ thus the current i _nB corresponding to sound waves emanating from the sound tube S _n of the final stage is determined. Here, S _1B and S _1A are coefficients expressed by the cross-sectional area ratio of the acoustic tubes adjacent to each other, and are expressed by equations (15) and (16), respectively.

S _1B = A ₁ / A ₁ + A ₂ …… (15) S _1A = A ₂ / A ₁ + A ₂ …… (16) A series of calculations of the current values of the blocks from the first stage to the last stage are instantaneous. Are executed, and these calculations are performed one after another at a predetermined timing. Here, the above (4)-
In equation (14), the value with the dash is the calculated value at the time t, and the value without the dash is the calculated value obtained by the calculation one time before the calculation at the time t. The digital value _inB obtained in this way is converted from digital to analog to produce an analog current, and this current is supplied to a speaker or the like to obtain audio.

Next, considering the propagation speed of the sound wave, this is similar to a transient phenomenon when an impulse is applied to an electric wire having a length 1 and an LC.

That is, as shown in FIG. 3, a line having LC is equivalently represented as shown in FIG. Here, the surge impedances Z ₀₁ and Z ₀₂ viewed from both ends are Becomes

Here, considering the traveling wave arriving from the other party as an equivalent current source, Next current if the intermediate is n delay circuit block Z _D, is output after n hours. That is, what has occurred in the left circuit has reached the right after τ hours.

I ₂ is the current on the feed pipe side Becomes However, in the digital calculation, since the voltage or the current is subdivided, V ₁ and V ₂ indicate the voltage at the measurement time t, and τ indicates the elapsed time.

In FIG. 4, if an impulse is applied to the L and C circuits, τ
After an hour, it goes to the output tube side. What arrived τ hours ago also represents that it has arrived at the partner. That the length l of the line 1 is easily calculated by one to normalize the delay block Z _D. l
The case engrave a 3cm may be a delay block Z _D 3 block.

Figure 2 (a) shows that the human vocal tract is about 17cm for males, so 1c
If the simulated acoustic tubes 17 present in increments m, the NyuTsuta waveform from A _1, if dividing the current half-cycle 10 the △ t and 10 .mu.sec, once emerges from A _n-side or 170Myusec.

It was but connexion, performed by the arithmetic processing apparatus the processing corresponding to the change in cross-sectional area A ₁ to A _n of the acoustic tube S ₁ to S _n, each part flowing individual equivalent circuit of the acoustic tube S ₁ to S _n current a memory having a value of impedance Z ₁ to Z _n corresponding to a ₁ to a _n necessary to calculate the values as a table, and calculation means for calculating a current value of each part of the equivalent circuit, and this equivalent circuit An arithmetic means for calculating the current value using the current values of the adjacent equivalent circuits to obtain an audio signal if the arithmetic processing is performed.
If D / A conversion is performed and output to a speaker, the sound is output from the speaker.

If an equivalent circuit is created by simulating 17 sound tubes in 1 cm increments as described above, it is easy to calculate as described above, and ordinary voices can be obtained, but voices closer to human voice (natural language) In order to obtain, the acoustic tube must be further subdivided to simulate a change in the cross-sectional area of the vocal tract corresponding to each natural language. However, in this case, the calculation becomes complicated and the memory of the arithmetic processing unit becomes enormous.

Therefore, we examined the mechanism of human vocalization and analyzed changes in the cross-sectional areas of the vocal tract, mouth, and lips. As a result, in all vocal sounds, the back of the vocal tract (the back of the throat) hardly changed.
It was found that the change increased as approaching the mouth, with the lip changing the most.

Focusing on this point, the present invention increases the number of memories by roughening the division of the acoustic tube corresponding to the back of the throat, and performing arithmetic processing on the division of the acoustic tube corresponding to the lips using a dense equivalent circuit. It is intended to obtain a voice closer to a human without performing.

FIG. 1 (A) shows a sound tube model diagram of the present invention corresponding to FIG. 2 (A), and FIG. 1 (A) shows its traveling wave equivalent model diagram. The same reference numerals are given or omitted.

Now, for example, in the case where the sound tube is divided into 17 sound tubes at 1 cm intervals and simulated, as shown in FIG. 1A, the sound tubes S _{1 to} S ₄ corresponding to the back of the throat are set at 4 cm intervals. It is regarded as one acoustic tube, and during this period it is controlled by the same data. Similarly 3cm the Supporting connexion S ₅ to S ₇ approaches the mouth, chopped 2cm to S ₈ to S _9, hid sequentially increments as hereinafter 1cm increments short sound tube S ₁₂ to S ₁₇ corresponding to the mouth and the lips The portion subdivides the acoustic tube corresponding to the portion where the cross-sectional area of the vocal tract changes greatly, for example, 1 cm, 0.5 cm or 0.2 cm.

Specifically, 1 chopped 1cm delay block Z _D described above
The normalized insert this delay block Z _D between the equivalent circuit of FIG. 1 (b) to the two acoustic tubes adjacent as shown. For increments 4cm inserting four delay blocks Z _D, and only to Zura computation is not carried out the timing of S ₂ to S ₄ performs calculation only corresponding to the acoustic tube S _1. Accordance connexion acoustic tube S ₂ ~S ₄
The memory required for the calculation is unnecessary. Similarly in the case of ticks 3cm Insert three delay blocks Z _D, also in the case of ticks 0.5cm accelerate arithmetic 1/2 or insert and speed. In this way, speech synthesis close to natural language can be obtained even if the number of variable cross-sectional area acoustic tubes near the throat is reduced, and the reduced amount near the throat can be increased toward the lips. For example, a voice closer to a natural language can be obtained without increasing the memory.

G. Effects of the Invention As described above, the present invention regards the vocal tract from the human vocal cords to the lips as a plurality of divided sound tubes having a variable cross-sectional area, and propagates two adjacent sound tubes of the divided sound tubes. Along with a current source and a parallel circuit having a surge impedance inversely proportional to the cross-sectional area of the acoustic tube, the length of the acoustic tube having the same area by having a delay circuit for propagation of the propagation current and making the constant of the delay circuit variable. The delay circuit constant is normalized to a unit that is easy to calculate and is blocked, and this delay block is provided more on the throat side where there is little change in the cross-sectional area at the time of voice generation, and the lip with a large change since so as to reduce insertion into the side, S ₁ required to calculate the current flowing through the respective portions of each of the equivalent circuit of the acoustic tube S ₁ to S _n ~
Without increasing the memory having a value of impedance Z ₁ to Z _n corresponding to S _n, the speech is obtained close to the natural language can be brought close to the change of the sectional area change of the human vocal tract of the acoustic tube.

[Brief description of the drawings]

1 is an acoustic tube and an electric circuit equivalent model diagram for explaining the present invention, FIG. 2 is an electric circuit equivalent model diagram of the acoustic tube, and FIG. 3 is an electric circuit diagram simulating sound propagation electrically. 4 shows an equivalent circuit diagram of FIG. S ₁ to S ₁₇ acoustic tubes ...... variable cross-sectional area, Z _D ...... delay circuit _{block, Z 1, Z 2, Z} n ...... surge impedance, Z _O ...... source impedance, Z _L ...... radiation impedance.

Claims (1)

(57) [Claims] 1. A voice synthesizer for calculating a current value based on a sound source wave model and a vocal tract model to synthesize a voice, wherein the vocal tract model includes an N-stage acoustic tube model and a plurality of delay blocks, The acoustic tube model of each stage calculates the current value based on the impedance corresponding to the cross-sectional area, the first acoustic tube model is connected to the sound source wave model, and the Nth acoustic tube model calculates the current value. A speech synthesizer characterized in that the current value obtained is supplied to a generation unit, and the number of delay blocks provided between acoustic tube models increases as the number of delay blocks approaches the sound source wave model side.