JPH0833752B2 - Speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION A. Field of Industrial Application The present invention relates to a speech synthesizer using an acoustic tube model.

B. SUMMARY OF THE INVENTION The present invention regards the human vocal tract as an acoustic tube group and associates it with a circuit element group having a surge impedance component, thereby simulating voice based on a current wave at the output end of the circuit element group. In the device to be created, the generation time of each phoneme is divided into multiple time zones for each phoneme that makes up the syllable, the phoneme parameters such as the cross-sectional area of the acoustic tube are specified for each time, and this phoneme parameter is specified. By performing the interpolation process according to the recurrence formula extracted from, the smooth voice that is close to the human voice is created.

C. Conventional technology Speech synthesis and music synthesizer (electronic musical instrument)
Recently, electronic devices that artificially synthesize and output so-called sounds such as LS have been used for LS for voice recognition and voice synthesis of one or several chips.
I has come to be realized at a low price by means of voice information processing and semiconductor large-scale integrated circuit technology, and various methods have been proposed depending on the purpose of use and constraints. In this speech synthesis, a raw voice generated by a human being is recorded, and the recording and editing method in which the raw voice is appropriately combined and edited into a sentence, and the human voice is not directly used. There is a parameter method in which only the parameter of (1) is extracted and the parameter is controlled to artificially generate a voice signal in the voice synthesis process.

In the parameter method, the voice waveform is sampled at a certain cycle, the value of the voice signal at each sampling point is converted from analog to digital, and the value is displayed as a code of 0 and 1. In order to perform faithful recording, it is necessary to increase the number of bits, which requires a large memory capacity.

Therefore, various highly efficient coding methods have been researched and developed in order to reduce the amount of this information as much as possible.

As one of the methods, there is a delta modulation method in which information of one audio signal corresponds to at least 1 bit. This method uses one bit to judge whether the next audio signal value is higher or lower than the current value, and if the value is higher, the code "1" is given, and if the value is lower, the code "0" is given. In the actual system configuration, a certain amplitude step amount (delta) is set, and the audio value obtained by the encoding so far and the input audio signal are stored so that errors are not accumulated. The residual signals of and are encoded.

This type of configuration predictive coding is called linear prediction method (predicting from several previous sample values) and Percoll method (a partial autocorrelation function called Percoll coefficient k is used instead of the prediction coefficient of the linear prediction method). There is.

D. Problems to be Solved by the Invention Among the conventional speech synthesis methods, the recording / editing method has a problem that the vocabulary and the types of sentences that can be synthesized are limited.

Further, in the method using the prediction code, it is difficult to combine articulatory sounds corresponding to the joints between sounds, and a combining unit method has not been established.For example, in utterances from vowels to consonants to vowels, vowel stationary The sound of the joint between the vowel and the vowel is cut off in the process from the transition of the vowel to the consonant and the transition of the vowel to the stationary sound of the vowel. Therefore, there is a problem that the sound is not smooth and does not give a natural feeling to humans.

An object of the present invention is to provide a speech synthesizer capable of synthesizing an arbitrary vocabulary and sentence, and having a smooth sound and being close to an actual human voice and giving a natural feeling to a listener. Especially.

E. Means and Actions for Solving Problems (1) Basic Concept A sound source is required to emit sound from the mouth,
This sound source is produced by the vocal cords. On the other hand, the vocal cords have the function of intermittently stopping exhalation by opening and closing two folds. Due to the intermittent flow, an air flow called a puff is produced, and when the vocal cords are tensioned, tension is applied to these folds and the frequency of opening and closing the folds. Becomes higher and a high frequency puff sound is generated. And when the expiratory flow is increased, a loud sound is produced.

When this sound source wave passes through a cylindrical acoustic tube such as the vocal tract, a certain component of the sound wave is emphasized by the resonance phenomenon from the open end, and a certain component is attenuated to form a complicated vowel waveform. The voice uttered by the mouth is affected by the shape of the vocal tract that passes through until it is radiated from the lips, even if the source waves have the same waveform. That is, the human generated sound is
It is determined by the length and cross-sectional area of the vocal tract from the vocal cords to the lips, and how the vocal cords tremble.

The present invention has been made paying attention to such a point, and the above vocal tract is regarded as a group of acoustic tubes having a plurality of variable cross-sectional areas, and a traveling wave phenomenon representing transmission of sound waves of the acoustic tube is equivalent to its equivalent circuit. The starting point is to realize by. When the vocal tract is regarded as an acoustic tube, the propagation of sound waves in each acoustic tube can be divided into a forward wave and a backward wave, which can be considered as repeated reflection and transmission phenomena at the boundary surface of each acoustic tube. And transmission are quantitatively defined according to the degree of mismatch of the acoustic characteristic impedance at the boundary surface, that is, the ratio of the cross-sectional areas of adjacent acoustic tubes. Here, the above-mentioned reflection and transmission phenomena are the same as the transient phenomena when an impulse current is passed through lines having different impedances in an electric circuit.

(2) Equivalent circuit For this reason, when an acoustic tube model composed of _n acoustic tubes S _{1 to} S _n is shown in FIG.
It can be represented as an electric circuit in which a circuit element group (T _{1 to} T _n ) composed of a lossless surge impedance component without resistance as shown in FIG. A _{1 to} A _n
Are cross-sectional areas of the acoustic tubes S _{1 to} S _n , respectively. In the present invention, basically, the above electric circuit is applied to change the impulse current supplied to the electric circuit and the surge impedance of each of the circuit elements T _{1 to} T _n to obtain the sound source wave of the acoustic tube model. Corresponding to changing the cross-sectional area of each acoustic tube, the current output from the circuit element T _{n at} the final stage is supplied to the voicing section such as a speaker to simulate the sound obtained from the acoustic tube model. Is producing.

Specifically, as shown in FIG. 1 (c), a circuit equivalent to the above electric circuit is assumed, the current of the current source in this equivalent circuit is changed with time, and an arithmetic expression as described later is given. Since the cross-sectional area ratio of the acoustic tube is introduced therein, each cross-sectional area A _{1 to} A _n is changed with time, and the current value of each part is calculated by this. P in the figure
Is a current source, Z ₀ is an impedance of the current source, Z _{1 to} Z _n are surge impedances of the circuit elements T _{1 to} T _n , Z _L is a radiation impedance, and i _{0A to} i _{(n-1) A} , i _1B to i _nB , a _0A ~
a _{(n-1) A} , a _{1B to} a _nB are currents in corresponding current paths indicated by symbols, W _{0A to} W _{(n-1) A} , W _{1B to} W _nB are current sources, and I _{0A to} 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, T _2, a current source W _1A made to correspond to the current I _1B flowing toward the middle circuitry T ₁ to T _2, the circuit elements T ₂ A current source W corresponding to a current I _1A flowing in the inside toward T _1.
Assuming a _1A, T current I _1B is at the boundary of the circuit element T _1, T _2
The reflected wave current i _1B reflected to ₁ and the transmitted wave current a _1A transmitted to T ₂ are separated, and the reflected wave current I _1A is reflected to T ₂ at the boundary between the circuit elements T ₂ and T _1. It is equivalently expressed that it is divided into i _1A and the transmitted wave current a _1B which is transmitted to T ₁ . Further, FIG. 4D is a schematic diagram schematically showing such a situation.

(3) Calculation First, the block including the first-stage current source P in FIG. 1C can be considered as a superposition of two circuits as shown in FIG. Therefore, assuming that the voltage of the current source P is V, the currents a ₁ and a ₂ in the figure are expressed by equations (1) and (2), respectively, and as a result, the current a _0A is expressed by equation (3).

a ₁ = V / Z ₀ + Z ₁ (1) a ₂ = Z ₀ / Z ₀ + Z ₁ · I ₀₁ (2) a _0A = a ₁ + a ₂ = 1 / Z ₀ + Z ₁ (V + Z ₀ · I _0A ) (3) Now, _assuming that current is supplied to the equivalent circuit for the first time, a _0A can be obtained by setting I _0A to zero. Then, the calculation is sequentially executed 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 figure as an example, they are expressed as the following equations (4) to (12).

a _0A ′ = 1 / Z ₀ + Z ₁ (V ′ + Z ₀ · I _0A ) (4) i _0A ′ = a _0A ′ −I _0A …… (5) I _0A ′ = i _1B ′ + a _1B …… (6) a _1B ′ = S _1B (I _1B + I _1A ) ... (7) i _1B ′ = a _1B ′ −I _1B …… (8) I _1B ′ = i _0A ′ + a _0A ′ …… (9) _{a 1A '= S 1A (I} 1B + I 1A) ...... (10) i 1A' = a 1A '-I 1B ...... (11) I 1A' = i 2B '+ a 2B' ...... (12) such As the calculation progresses, the equations for the final block are expressed as equations (13) to (15).

a _nB ′ = Z _L / Z _n + Z _L · I _nB …… (13) i _nB ′ = a _nB ′ −I _nB I _nB ′ = i _{(n-1) A} + a _{(n-1) A} ...... ( 14) Thus, the current i _nB corresponding to the sound wave emitted from the final stage acoustic tube S _n is obtained. However, S _1B and S _1A are coefficients represented by the cross-sectional area ratios of the acoustic tubes adjacent to each other, and are represented by the 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 calculation of the current value of the block from the first stage to the final stage is instantaneous It is executed and these calculations are performed one after another at a predetermined timing. Here (4)-
In the equation (14), a value with a dash is a calculated value at the time t, and a value without a dash is a calculated value obtained by the calculation one time before the calculation at the time t. The digital value i _nB thus obtained is _subjected to digital / analog conversion to create an analog current, and a voice is obtained by supplying this current to a speaker or the like. The timing of the calculation is determined in consideration of the speed of sound and, for example, by setting the propagation time of one of the acoustic tubes as the time interval of the calculation, the backward wave currents I _{0A to} I _{(n-1) A} and the forward wave current are forwarded. Wave current I _1B ~ I
_nB is created a state equivalent to a state flowing in each circuit element T ₁ ~I _n A at the same speed as the speed of sound, and thereby align the acoustic tube model and the electric circuit model.

The present invention is based on the realization of the equivalent model and calculation as described above. Specifically, the generation time of each phoneme is divided into a plurality of time zones for each phoneme that constitutes a syllable, and each time for each band, a repetition frequency of the sound source wave pitch, each initial value X _r of the next time period from the initial values Xo for the initial value and the time zone of the cross-sectional area of the energy and acoustic tube of the sound wave
A phoneme parameter storage unit that stores a time constant and a sound source wave pattern that defines the manner of change to, a parameter interpolation processing unit that performs interpolation processing of the pitch, energy, and cross-sectional area corresponding to the input phoneme data, An arithmetic unit for calculating a current value output from the output end of the circuit element group based on the parameter interpolated here, and a voicing unit for generating a voice based on the arithmetic result of the arithmetic unit, parameter interpolation processing unit performs a number of times the interpolation calculation using said X _r the equivalent to the initial value Xo and the target value during each time period, the interpolation operation, the n-th interpolation operation value X
(N), when the time constant is represented by D, it is executed according to a recurrence formula represented by X (n) = D { _Xr- X (n-1)} + X (n-1). It was done.

F. Embodiment FIG. 1 is a diagram showing a block configuration of an embodiment of the present invention. Reference numeral 1 denotes a Japanese language processing unit, which performs illegible conversion and the like on the input Japanese sentence by referring to a segment break or a dictionary. Reference numeral 2 denotes a text processing unit, which performs processing to add intonation to text. A syllable processing unit 3 attaches an accent according to the intonation to the syllables forming the sentence. For example, the sentence "Sakura ga sai" is decomposed into syllables such as "SA", "KU", "RA", etc., and each syllable is accented. Since the intonation of a sound is determined by the repetition frequency of a sound source wave, its energy and time, which will be described later, adding an accent means determining the coefficient for these parameters. 4 phoneme processor, 4 ₁ are syllable parameter storage unit, the phoneme processor 4 is input to the "SA" ... etc. syllable data, correspondence between the phoneme is a unit of syllable and vowels and consonants Processing for decomposing into phonemes by referring to the data in the syllabic parameter storage unit 4 ₁ which defines, for example, the phonemes "S" and "A" are extracted for the syllable "SA".

Reference numeral 5 is a parameter interpolation processing unit, 5 ₁ is a phoneme parameter storage unit, and 5 ₂ is a sound source parameter storage unit. As shown in FIG. 4, the phoneme parameter storage unit 5 ₁ divides the utterance time of each phoneme into a plurality of, for example, three time zones 0 _{1 to} 0 ₃ , and a pitch that is a repetition frequency of a continuous time source wave for each time zone. , The initial value of the energy of this sound source wave and the cross-sectional area of the acoustic tube, and the time constant and the sound source wave pattern that specify the method of changing from the initial value of the time zone to the initial value of the next time zone. Is stored. In this example, the human vocal tract (about 17 cm for a male) is modeled by connecting 17 acoustic tubes each having a length of 1 cm. Therefore, the cross-sectional area value per one time zone is calculated.
17 pieces (A _{1 to} A ₁₇ ) are defined. Also the tone generator parameter storage unit 5 _2, for example, three waveform components of the sound source wave pattern G ₁ ~G ₃ as shown in Figure 5 are stored as 50 samples data. The parameter interpolation processing unit 5 is a part that performs an interpolation process of the pitch, energy, and cross-sectional area in each time zone (O _{1 to} O ₃ ), and this processing is each parameter of the pitch, energy, and cross-sectional area in the relevant time zone. Let Xo be the initial value of X, the initial value in the next time zone be X _r , the nth interpolation calculation value be X (n), and the time constant corresponding to each parameter be D. This is a process of performing calculation n times during the time period according to the recurrence formula shown. However, the initial value X (O) is the above Xo.

X (n) = D { _Xr- X (n-1)} + X (n-1) ...
(17) For example, in the pitch interpolation process in the time zone O ₁ , since Xo corresponds to P ₁ and X _r corresponds to P ₂ , calculation is performed according to the equation (18).

_{X (n) = DP 1 {} P 2 -X (n-1)} + X (n-1) ...
(18) Here, the above equation (17) is a recurrence equation of the following equation (19).

X = X _r −e ^−Dt (19) That is, when the equation (19) is differentiated, the equation (20) is established, and thus the equation (21) is established.

dx / dt = De ^-Dt (20) ΔX = X (n-1) -X (n) = Δt · De ^{-Dt (n)} = ^ΔtD (X _r −X (n)) …… ( 21) Therefore, formula (22) is obtained.

X (n + 1) = Δt · D (X _r −X (n)) + X (n)
(22) Here, since the time interval of interpolation calculation is constant, Δt · D
Can be collectively replaced with the time constant D, which is expressed as the equation (17).

Reference numeral 6 denotes an arithmetic unit, which obtains a digital value of the current _inB shown in FIG. 1C at a time interval of 100 μs, for example, at the same timing as the interpolation calculation based on the parameters calculated by the parameter interpolation processing unit 5. Reference numeral 7 denotes a digital / analog (D / A) converter, which produces a current wave (analog current) based on the digital value obtained by the calculation unit 6. Reference numeral 8 is a voicing section such as a speaker, which utters a voice based on an analog current.

 Next, the operation of the above embodiment will be described.

Japanese sentences input by word processor etc.
After passing through the Japanese language processing unit 1, the sentence processing unit 2, and the syllable processing unit 3, intonation and the like are added to divide into syllable units,
Furthermore, the phoneme processing unit 4 decomposes each syllable into phonemes.
Next, the parameter interpolation processing unit calculates the pitch, energy, and cross-sectional area of each phoneme in the phoneme parameter storage unit 5 ₁
Retrieved from, for these parameters each time period (0 _{1-0 3)} interpolation processing for each is performed. This interpolation processing is performed according to the equation (17), and for example, the pitch in the time zone 0 ₁ is performed according to the equation (18). FIG. 6 is a diagram showing this state, in which the respective interpolated values P (1), P (2) ... P (n) of the pitch obtained by the interpolating operation follow the curve expressed by the following equation (23). Will be lined up.

^{_{P = P 2 -e -Dt ...... (}} 23) The sample data of the sound source wave pattern defined for each time slot 0 _{1-0 3} is taken out from the sound source parameter storage unit 5 _2, the sample data and the pitch and the like The interpolated value of 1 is given to the arithmetic unit 6, and the arithmetic unit 6 executes the arithmetic operation described in detail in the above-mentioned E. (3) item "Operation". In this calculation, the syllable processing unit 3 multiplies the coefficient or function corresponding to the accent attached to each syllable unit and each parameter obtained by the parameter interpolation processing unit 5 so that the intonation of the sentence appears. Is calculated. In this way, the digital value of the current wave corresponding to the sound wave emitted from the final stage acoustic tube is obtained, the current wave is generated by the D / A converter 7 based on this value, and the corresponding sound is emitted from the vocal sound 8. .

G. Effects of the Invention According to the present invention, the sound wave propagation of the acoustic tube model is replaced by the current flow of the equivalent circuit, and parameters such as the pitch of the current source and the cross-sectional area of the acoustic tube are defined for each phoneme. Interpolation of parameters based on an exponential function is executed for a seam between seams or a seam of a time zone divided in a phoneme, so that a smooth voice can be obtained and a listener feels natural. The interpolation processing of the exponential function is not actually performed, but each interpolation value is obtained by using the recurrence formula extracted from this function. Further, it is sufficient to store data in units of phonemes or in units of time zones instead of storing all the parameter values in the area corresponding to the joint between phonemes in the memory, so that the memory capacity can be small.

[Brief description of drawings]

1 is an explanatory view showing an equivalent model of an acoustic tube, FIG. 2 is an equivalent circuit diagram showing a block including a current source, FIG. 3 is a block diagram showing an embodiment of the present invention, and FIG. 4 is a phoneme parameter. Data diagram, FIG. 5 is an explanatory diagram showing a sound source wave pattern, FIG.
The figure is an explanatory diagram showing how parameter interpolation processing is performed. 4 ...... phoneme processor, 4 ₁ ...... syllable parameter storage unit, 5
...... Parameter interpolation processing unit, 5 ₁ ...... Phoneme parameter storage unit, 5 ₂ ...... Sound source wave pattern storage unit, 6 ...... Calculation unit,
7 ... Digital / analog converter, 8 ... speaker.

Claims (1)

[Claims] 1. A human vocal tract is regarded as a plurality of acoustic tubes joined in cascade, and a group of these acoustic tubes and a circuit element group of a surge impedance component are made to correspond to each other, and an audio source and a current source are made to correspond to each other. In a speech synthesizer that imitates the sound wave emitted from the output end of the acoustic tube group based on the current wave at the output end of the circuit element group, the utterance time of each phoneme is set to 1 or more for each phoneme that constitutes a syllable. It is divided into time zones, and for each time zone, the pitch, which is the repetition frequency of the sound source wave, each initial value of the energy of this sound source wave and the cross-sectional area of the acoustic tube, and the following time from each initial value Xo of the relevant time zone. A phoneme parameter storage unit that stores a time constant and a sound source wave pattern that defines how the band changes to each initial value X _r , and each interpolation process of the pitch, energy, and cross-sectional area corresponding to the input phoneme data. To A parameter interpolation processing section, a calculation section for calculating a current value output from the output end of the circuit element group based on the parameter interpolated here, and a sound is generated based on the calculation result of this calculation section. and a vocal portion, the parameter interpolation process unit performs a number of times the interpolation calculation using said X _r the equivalent to the initial value Xo and the target value during each time period, the interpolation calculation, n The recurrence formula represented by X (n) = D { _Xr- X (n-1)} + X (n-1), where X (n) is the th interpolation calculation value and D is the time constant. A speech synthesizer characterized by being executed according to the following.