JPH01219896A - Speech synthesis system - Google Patents

Abstract

PURPOSE:To obtain a smooth voice by extracting information on a specific exponential function contained in a table at extraction intervals corresponding to the processing result of a speech processing part and determining execution parameters, and calculating the current value when the signal of a sound source is supplied to an acoustic tube model according to the parameters and generating a speech synthesis signal. CONSTITUTION:An interpolation processing part 6 extracts the exponential function information stored in a memory 5 at the extraction intervals corresponding to the phoneme time constant, pitch time constant, and energy time constant of each sectioned phoneme. Then the exponential function information is obtained by extending the exponent curve in the memory 5 according to the information extraction intervals in the time-base direction. The signal of the sound source is supplied to the acoustic tube equivalent circuit 8 according to the execution parameters and an arithmetic part 9 calculates the value of the current which flows at this time. Thus, respective parameters are interpolated by phonemes which are sectioned into a specific section, so the smooth voice having no break is obtained.

Description

[Detailed description of the invention] A. Industrial application field The present invention relates to a speech synthesis method.

B0 Summary of the Invention The present invention is a method for simulating speech synthesis using an acoustic tube whose surge impedance changes as the cross-sectional area changes. By extracting at a certain extraction interval and determining execution parameters, and calculating the current value when the sound source signal is supplied to the acoustic tube model based on the determined execution parameters, a smooth speech synthesis signal can be obtained. This makes it possible to obtain audio, reduce the amount of information required for interpolation processing, and eliminate the need to perform complex interpolation operations such as multiplication.

C1 Conventional technology Electronic devices that artificially synthesize and output speech have recently become low-cost LSIs for speech recognition and speech synthesis of one or several chips thanks to speech information processing and semiconductor large-scale integrated circuit technology. Various methods have been proposed depending on the purpose of use and constraints. For this speech synthesis,
There are recording and editing methods that record raw human voices and combine them appropriately to edit them into sentences, and two methods that do not use the human voice directly but extract only the parameters of the human voice. There is a method of artificially creating a speech signal by controlling the parameters during the speech synthesis process.

There is a PARCOR method which is widely used because it allows high-quality synthesized sounds to be obtained using this parameter method.

When handling audio using an electronic computer, the audio waveform is sampled at certain intervals, the audio signal value at each sampling point is converted from analog to digital, and the resulting values are displayed as codes of 0 and 1. To record faithfully to analog signals, it is necessary to increase the number of bits, but voice synthesis signals require a large amount of memory.

Therefore, various highly efficient encoding methods are being researched and developed in order to reduce the amount of information as much as possible.

As one of the methods, for information of one audio signal,
There is a delta modulation method, which uses a minimum of 1 bit. This method uses one bit to determine whether the next audio signal value is higher or lower than the current value, and if it is high, it gives the code "l" and if it is low, it gives the code "0" to encode the audio signal. In the actual system configuration, a constant amplitude step flaw (delta) is determined, and the audio value obtained by conventional encoding is compared with the input audio signal to prevent errors from accumulating. The residual signal is encoded.

This kind of configuration is called predictive coding, and is based on the linear prediction method (
There are two methods: the Percoll method (which uses a partial autocorrelation function called a Percoll coefficient instead of the prediction coefficient of the linear prediction method).

D8 Problems to be Solved by the Invention As mentioned above, the method using predictive coding has a problem in that it is difficult to make articulatory connections, which correspond to joints between sounds.

For example, in the process of vocalization from a vowel to a consonant to a vowel, the sound at the joint between the vowels is cut off during the process from the steady vowel to the transient to the consonant, and the transitional vowel to the steady vowel. It doesn't give a natural feel when you use it.

Furthermore, in the case of musical instrument sound synthesis, the joints between scales are important, but since the synthesis method differs from the principle of sound generation in an actual musical instrument, it still lacks a natural feel, especially in reverberant sounds. In order to approximate natural sounds in both of these systems, there are problems such as the need for a large number of electronic components such as memory and arithmetic units, making the device expensive.

Furthermore, when performing interpolation processing so that the seams between sounds do not break, a problem arises in that calculations for the interpolation processing become complicated.

Therefore, the inventors of this application believe that since the generation of human sounds and the musical sounds of musical instruments are created by changes in the shape of the human oral cavity and the length and cross-sectional area of acoustic tubes, the traveling wave phenomenon representing the transmission of sound waves in these acoustic tubes is Analyzed using a tube equivalent circuit, focusing on the fact that the cross-sectional area of the acoustic tube is inversely proportional to the surge impedance, we simulated the cross-sectional area by changing the surge impedance, and achieved articulatory coupling by continuously changing the surge impedance. He devised a speech synthesis method that facilitates the synthesis of sounds similar to human speech and improves the naturalness of speech, and filed a patent application for this method (patent application). 62-91705, hereinafter referred to as the earlier application).

The present invention has been made in view of the above points, and its purpose is to make it possible to obtain easy-to-hear and smooth speech based on the invention of the earlier application, and to reduce the amount of information necessary for interpolation processing. It is an object of the present invention to provide a speech synthesis method that reduces the number of operations and eliminates the need for complicated operations.

E9 Means for Solving Problems The present invention is a speech synthesis method that uses phonemes as a basic unit and simulates them using an acoustic tube, in which l phonemes are divided into a predetermined number and treated, and a built-in language dictionary is referred to. a language processing unit that performs linguistic analysis of input character data; and a speech processing unit that has syllable parameters and the classified phoneme parameters and performs syllable processing and phoneme processing of the data analyzed by the language processing unit. and a memory in which information about an exponential function expressed by a predetermined exponential curve is stored, and the execution parameters are determined by extracting the information in the memory at an extraction interval corresponding to the processing result of the audio processing section. and an interpolation processing section that controls the current flowing through the equivalent circuit of the acoustic tube, to which the sound source signal is supplied and whose surge impedance changes as the cross-sectional area changes, to a predetermined value based on the execution parameters determined by the interpolation processing section. The present invention is characterized in that it includes a calculation section that performs calculations based on timing, and obtains a speech synthesis signal based on the calculation results of the calculation section.

The input character data is subjected to language analysis in the language processing section, and then subjected to syllable processing and phoneme processing in the speech processing section. The data obtained by the audio processing section is processed by an interpolation processing section for each phoneme divided into a predetermined number, such as acoustic tube cross section, phoneme time constant, duration, pitch, pitch time constant, energy, energy time constant, sound source, etc. Interpolation processing is performed for each parameter. The interpolation processing section extracts the exponential function information stored in the memory at extraction intervals corresponding to the phoneme time constant, pitch time constant, and energy time constant for each of the classified phonemes. Then, exponential function information obtained by expanding the exponential curve in the memory in the time axis direction according to the size of the information extraction interval is obtained as the output of the interpolation processing section. Since the extraction interval is changed according to each time constant, exponential function information can be obtained according to the time constant for each phoneme divided into a predetermined number. In this way, in the interpolation processing section, information (
Exponential curves with various slopes can be obtained from (information in memory). This allows information on the exponential function used for interpolation processing (exponential function information with different time constants)
There is no need to store a large number of. Furthermore, since it is sufficient to simply change the information extraction interval in accordance with the size of the time constant, complex operations such as multiplication are not necessary. Each execution parameter for input character data is determined by the operation of the interpolation processing section. Based on this execution parameter, the sound source signal is supplied to the acoustic tube equivalent circuit, and the current value flowing at that time is
The calculation unit performs calculations at timings that match the speed at which the current travels. By D/A converting this calculation output and supplying it to a speaker, a speech synthesis signal corresponding to the input character data can be obtained. Since parameters are interpolated for each phoneme divided into a predetermined number in this way, smooth speech without any traces can be obtained.

G. Embodiment Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

First, the invention of the earlier application, which is the basis of the present application, will be explained in detail.

The change in the cross-sectional area of the vocal tract during speech production is caused by, for example, when making the ``a'' sound, the back of the throat is narrow and the lips are open, and the exhaled air that is forced out of the lungs causes the vocal cords to open and close intermittently to absorb the exhaled air. The sound wave that is repeatedly reflected in the road (acoustic tube) comes out as the sound waveform of "A". When the throat is wider and the tip of the lips are narrower, the sound waveform of "i" is output.

In this way, the frequency is determined by the shape of the mouth, and if the shape of the mouth is simulated, "a" or "i" will be uttered.

The shape of the mouth can be simulated by the cross-sectional area of the sound tube, and changes in the cross-sectional area of the sound tube can be simulated by changes in surge admittance. Therefore, the shape of the mouth can be simulated by changing the surge admittance. Since changes in surge admittance can be varied very easily in electrical circuits, various sounds can be synthesized using electrical signals. FIG. 9(a) shows a simulation of the vocal tract by connecting acoustic tubes with different cross-sectional areas A, At...80. In the same figure (A), the acoustic impedance is replaced with an LC circuit of an electric circuit.
Each sound tube is treated as one LC line, and the whole is a concentrated line n-
1 electric circuit. Figure 9 (1) is a traveling wave equivalent model diagram, and the acoustic impedance of each acoustic tube is 2.
．． ．． 2. ...Z,% is inversely proportional to the cross-sectional area of the acoustic tube (,
The acoustic admittance is proportional) and is proportional to the speed of the sound wave, so A IA t A n . In addition, in the same figure, Zll is the sound source impedance, and Z
L indicates radiation impedance, and arrows between blocks indicate forward waves and backward waves.

If you want to make the sound "A" now, give the shape of the mouth of "A" at the cross-sectional area of the acoustic tube corresponding to the tip of the lips in Figure 10, and apply the impulse P intermittently. , the sound "a" is obtained, and if you want to produce the sound "i" from "a", the cross-sectional area of the acoustic tube corresponding to the tip of the lips is narrowed, as shown in the same figure (a), and the "i" sound is produced. ``I'' can be obtained by giving the appearance of the mouth.

When the impulse P is applied continuously and intermittently to change the entire cross-sectional area to resemble the mouth of "i", the vocal tract is simulated by n acoustic tubes shown in Figure 9, so these By moving each cross-sectional area from "A" to "A", the shape of the mouth will change continuously to "A". Changing the cross-sectional area of the acoustic tube is done by gradually changing the surge impedance.

Therefore, since the cross-sectional area can be changed continuously, it is of course possible to obtain the steady-state sounds "A" and "IJ," but since the impedance can also be changed continuously, the sounds in between, i.e. It is possible to obtain sounds between sounds.Therefore, articulatory combination similar to human pronunciation is performed smoothly without any sound breaks.

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

That is, the line having the LC as shown in FIG. 11 is equivalently represented as shown in FIG. 12. Here is the surge impedance Z seen from both ends. 1. Zo2 becomes Zo--rτ, Zoz=q.

Here, if we consider the traveling wave that has arrived from the other party as an equivalent current source, the current will be output after n hours if there are n delay circuit blocks Z in the middle. In other words, what is generated in the left circuit reaches the right circuit after τ time. However, in digital calculation, since the voltage or current is subdivided, V, , v represents the voltage at measurement time t, and τ represents the elapsed time.

In Figure 9, if an impulse is applied to the L and C circuits, τ
After some time, it will come out to the output tube side. This equivalently represents that what has reached the τ time plane has also reached the other party. Setting the line length Q to 1 can be easily calculated by normalizing the delay block n to I. Q
When dividing into 3 cm, n of the delay block should be set to 3 blocks.

The human vocal tract is approximately 17cm+ in males, so Figure 9 (a)
If we simulate the acoustic tube with 17 acoustic tubes in Ic0I increments, the waveform that enters from A will take 170 μsec to exit from the Arl side if we divide the half-cycle current into IO and set the Δt to 10 μsec. come.

Next, an example will be described in which speech is synthesized based on character input signals using the acoustic tube model described above. In Fig. 1, 1 is the Japanese language processing unit, which receives a sentence containing kanji and kana as input, matches it with dictionary 2, and processes it into phrase 1.
Natural language analysis is performed to divide nine phrases and classify morphemes, and furthermore, accent processing is performed, phonetic conversion is performed, and intonation is added to create sentence processing data. 3 of the sound processing means 12 is a syllable processing section, which has syllable parameters as shown in FIG. 2, and performs syllable processing on the data created by the Japanese processing section 1. The syllable parameter is 110 of the consonant.
For each syllable, the pitch (pitch), energy (
Give the sound intensity) and duration, and adjust these values to match the audio level. For example, in the case of "cherry blossoms", the pitch P for each syllable of SA, KU, and RA is shown in Figure 2.
1 energy E1 time T is normalized.

4 is a phoneme processing unit which has phoneme parameters with a parameter interpolation function. The phoneme parameters are divided into the rising part Or, the steady part Ot, and the falling part 03 for each phoneme, and the phoneme (cross-sectional area) time constant, I! Duration time pitch, pitch time constant.

Normalize the energy, energy time constant, and sound source to form a block of data for each section. Taking the above-mentioned "cherry blossom" as an example, it is "S" as shown in Figure 3.

For each phoneme rAJ, rKJ, rUJ, rRJ, rAJ,
Rising part of the division 0. If so, Do.

'rl, P+, DP+, E+, DE+, G1 data units are formed. These data units correspond to the cross-sectional areas A, ~A, of the acoustic tube model in Figure 9, each having a cross-sectional area of A0.1 ~
They are provided corresponding to An and n. In other words, if the cross-sectional area A of the acoustic tube model is 17, then for each syllable, 6x17=
Data of I02 is prepared.

Each of the above time constants specifies how to move from the final value of the previous segment to the target value corresponding to each of that segment. Time T is a duration time, and each of the above processes is performed within this time T. Further, the sound sources G + , G t , and G s provide sound sources in the form of 300 pulses within time T for each segment in the consonant part, and about 50 pulses in the vowel part.

The sound source uses an impulse signal as a vowel sound source, for example,
A white noise signal is used as a consonant source, for example, one extracted from human speech, and the sound source data is shown, for example, as shown in FIG.

All the data of each parameter mentioned above is stored in the memory 5. 6 is an interpolation processing unit that performs interpolation processing on the parameters and determines execution parameters. The interpolation processing unit 6 extracts one piece of exponential function information previously stored in an internal memory (not shown) at an extraction interval corresponding to the time constant obtained by the phoneme processing unit 4, and thereby extracts the following ( +), (2), and (3) to obtain information equivalent to when calculating the interpolation calculation formulas.

The pitch interpolation calculation formula is P (--11-DP (P t-+ P (.)+
P, ...... (1) (DP is the pitch time constant, P (r+ is the target pitch value, and the step of n is, for example, every 100 μs.) Energy interpolation calculation formula is Et−−++=DE (Elo Et−u) 10Et1
1+...River...(2) (However, DE is the energy time constant, E(r) is the target value of energy, and the increment of n is, for example, every 100 μs.) The interpolation calculation formula for the cross-sectional area is , All, l-11-DS (All-+ All-+
) +At+-+--(3) (However, DS is the area time constant, Ailr+ is the target value of the cross-sectional area, and the increments of n are, for example, every 400 μs.) That is, in the memory (not shown) of the interpolation processing unit 6. Figure 5 (a
), (b), (c), and (d) are stored, and the time constants (DP, DE in FIG. 3) of each phoneme obtained by the phoneme processing unit 4 are stored.

The information on the exponential curve is extracted at an extraction interval corresponding to Do1 (where Do is equivalent to DS). For this reason (1), (2)
, (3), information corresponding to the magnitude of the time constant (equivalent to that obtained when the calculation is performed) can be obtained. The basis for this is shown in Figure 5(e) below.

This is explained in conjunction with (f). First, the exponential function is y=(1-e-
Dl)・xl・・・・・・・・・(4), and its recurrence formula is 'I n*+= D (x r x n) + X
n...(5) (i.e. (1), (2),
(3) P l'+61+ Ene+, A ln* of the formula
Replace + with Y no+, DP.

Replace DE and DS with D, replace PF, E, , At, with Xr, and replace Pn, En, A+n with X. ). If we set x=D−t from the above equation (4), the value of X uniquely determines the target value when a certain time constant is counted for one hour. Therefore, if we consider that X = constant, the reached value at time t is D
However, the value at this time is obtained by expanding or contracting the curve in FIG. 5(e) in the time axis direction like the curve in FIG. 5(f).

Therefore, we created one table as shown in Figure 5(e) and
For example, if D is large, skip reading will be increased when reading from the table, and if D is small, skip reading will be decreased.
FIG. 5(f)) can be freely obtained. This results in the above (
5) There is no need to calculate equations, making it possible to simplify and speed up interpolation processing. Furthermore, there is no need to have separate tables of exponential functions depending on the size of the time constant, and the memory capacity in the interpolation processing section 6 can be extremely small.

Note that the interpolation calculation formulas (1), (2), and (3) are y=1
, -Dt . . . -(6), and the method of derivation will be explained below. Now, as shown in Fig. 6, the time tn
The slope when transitioning from to time t1+Δt is the above-mentioned (6th
), t (where D is a time constant) is obtained. Therefore, if At is a time step, the value 3'n of time tn and the value yno1 of time t0+Δt are y, 1
=le-e-Dt″ ・・・・・・・・・(8)y,,,
=l e→(tn*Δ1 = 1-e-Djn, e-DΔ' ・------
--C9). Since the slope at time tn is t, Δy, that is, 'in-+-yn, is
From formula 6), Y nht Y n=Δt@De-Dtn=Δt-D
It is expressed as (1-y,)-(11). Therefore, if we shift the equation (11) above, we get yn*+−Δt−D (1−yn) −1−y, −=・
The formula (12) is derived, where Δt=1.
If the target value is yr, Yn-t=D (Yt Yn) +-yrl...
...(13).

Reference numeral 8 in the sound synthesis means 13 is an acoustic tube model section, to which the signal of the sound source P is supplied as shown in FIG. 9, and the cross-sectional area A, ~
An acoustic equivalent circuit of a sound tube model formed by connecting sound tubes of An is used. Reference numeral 9 denotes a calculation unit that calculates at a predetermined timing, based on the execution parameters determined by the interpolation processing unit 6, the current that flows when the signal of the sound source P is supplied to the sound tube mocker. Note that the calculation section 9 is realized, for example, as described later, by a calculation processing device comprising a memory having a table of current values and functions.

Using the execution parameters determined by the interpolation processing unit 6 as described above, the sound source signal is input to the acoustic tube model and the cross-sectional area change of the acoustic tube is simulated, and the current value obtained by calculation is the voice synthesis waveform. It is input to the section IO. This speech synthesis waveform part! 0 converts the input digital signal into an analog signal and outputs it as audio from the speaker 11.

Next, the operation of the above embodiment will be explained with reference to the flowcharts of FIGS. 7 and 8. FIG. 7 shows the overall software, and FIG. 8 shows the software of the calculation section 9. First, in steps S and -S in Fig. 7, the Japanese processing unit l manually receives a sentence containing kanji and kana, matches it with the dictionary 2, delimits each phrase and sentence, and performs natural language analysis for morphological classification. Then, perform accent processing, convert this into phonetic form, add intonation, and create sentence processing data. Next, the syllable processing unit 3 performs syllable processing on the data created by the Japanese language processing unit 1 in step S@, and the phoneme processing unit 4 performs phoneme processing in step S. Next, in step S, the interpolation processing unit 6 interpolates the phoneme level data of the phoneme parameters for each category as shown in FIG. 3 to determine each execution parameter. That is, as described above, the exponential function information stored in the internal memory in advance is extracted at an extraction interval corresponding to the time constant obtained by the phoneme processing section 4, and the above (1),
Information equivalent to when calculating the interpolation calculation formulas (2) and (3) is obtained. This allows you to change the way the sound source is given (each section 0
．． , os, 03) and each time constant (how to move from the final value of the previous section to the target value of the section) are specified. Next, the calculation unit 9
In step 8, the current when the signal of the sound source P is supplied to the acoustic tube model is calculated and determined at a predetermined timing using the determined execution parameters.

The calculation ff9 is composed of an arithmetic processing device such as a computer, and calculates the current value, and the processing is performed by the eighth
This is done according to the flowchart shown in the figure. Here, the relationship between the calculation items and calculation formulas of the calculation unit 9, the acoustic tube model, and the equivalent circuit is shown in FIGS. do. First, when an impulse is input to the acoustic tube AI, the arithmetic processing unit consisting of a computer reads A from the memory in step Sl,
Extract aox, ioA, IOA, E of (
a OA ""' a n-lA is the passing wave current, ■OA
-j n-IA is reflected wave current, and ■i~I n-IA is backward wave current).

Based on the extracted value, in step S, aoA′=
f (E, IoA)i 0A' = aoA'
Perform the calculation of I OA . This calculated value a. A'+ I OA'
and the value aha, aI^, its, lIA, 11B, I+^(a
tB~ans is the passing wave current, l ta ""' 1 n
B indicates a reflected wave current, and I + B' = 1 nB indicates a forward wave current. Note that aOA indicates the previous value, and aOA' indicates the current value, and the reflected wave current, backward wave current, and forward wave current are also expressed in the same manner. ), and in step S4, a + a' = S I[1(I IB
+ I IA) a IA' = S IA (+ +
e + I IA) i IB' = a IR'
118iIA' = a+^' I IB 11B' = l OA'a OA' is calculated. In step S, l obtained in S4
10A' = i using +a' and a +n'
Calculate lE1+ a +vs. On the other hand, the value i IA', a1^' obtained at S, and the value a tB, a of the tube A3 introduced from the memory in step S6.
tA+l tB+1! At I je, I
The following calculation is performed in step S7 using the following.

a 2B+ = S ta (I tB+ 1 tA)
a! ＾′ :S! ^(■tB+ l tA)i to'
= awe'I! B ixA' = atA'ItA 12o' = 11A + a IA' In step S8, it, 3' obtained in S7.

Using i', the calculation 11A'=lyB'+a2A' is performed. Below, the cross-sectional area A1 of the acoustic tube simulated in the same manner
Calculations corresponding to An are performed, and step S n-+
Then, ans' -f (I n5) l nB'= a nB' I nBI n5-j
Perform the calculation n-IA+a rl-IA. In step S, the result is used to calculate In-IA'=1 nB'+a nB'. That is, the output of the calculation result at the final stage (n stage) of the equivalent circuit corresponding to A1 to An of the acoustic tube is D/A converted by the speech synthesis waveform unit 10 and output to the speaker 11, and is output as audio from the speaker 11. be done.

Therefore, since the arithmetic processing device performs arithmetic operations corresponding to the acoustic tubes A1 to An, this arithmetic processing device calculates the current value of each part flowing through the equivalent circuit of each acoustic tube A, to An, and the function f, SIB, 5IA (i=1,2...
- a memory having n-1) as a table, and a first calculation means for calculating the current value of each part of the equivalent circuit;
The equivalent circuit is provided with second calculation means for calculating the current value of the equivalent circuit using the current value of the equivalent circuit adjacent to the equivalent circuit.

The timing of the calculation corresponding to each acoustic tube A, to A1 is adjusted to the time Δt for the current to reach the adjacent acoustic tube. That is, if the current in the adjacent acoustic tube is, for example, l
If it is reached after Oμs, then the calculations corresponding to the acoustic tubes A, ~An are performed sequentially at time intervals of 10μs.

In the embodiment described above, the summer phoneme is divided into three parts: a rising part, a steady part, and a falling part.

1-1. Effects of the Invention As described above, according to the present invention, the following effects can be obtained.

(1) Since each parameter is interpolated for each phoneme divided into a predetermined number, a smooth speech synthesis signal without any traces can be obtained.

(2) Execution parameters are determined simply by extracting information on a predetermined exponential function at an extraction interval corresponding to the processing results of the audio processing unit, eliminating the need for complex operations such as multiplication, simplifying interpolation processing and Speed-up can be achieved.

(3) It is not necessary to have information on each exponential function depending on the size of the time constant, and the amount of information required for interpolation processing can be extremely small. Therefore, the capacity of the memory can be reduced.

(4) Since a parameter is provided for each segmented phoneme and interpolation processing is performed, there is no need to provide all the information required to reach the target value, and the memory capacity can be small.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is an explanatory diagram of syllable parameters, Fig. 3 is an explanatory diagram of phoneme parameters, Fig. 4 is a signal waveform diagram of sound source data, and Fig. 5 (a). ,(
b), (c), (d). (e) and (f) are curve diagrams of exponential functions, Figure 6 is an explanatory diagram for deriving the recurrence formula, Figure 7 is a flowchart for explaining the overall configuration of one embodiment, and Figure 8 is Flowchart of the calculation section, FIG. 9 is an explanatory diagram of the acoustic tube model section and the calculation section, FIG. 10 is an explanatory diagram of changes in the vocal tract, FIG. 11 is an electric circuit diagram electrically simulating sound propagation, and FIG. 12 is an equivalent circuit diagram of FIG. 11. l... Japanese language processing unit, 2... Dictionary, 3... Syllable processing unit, 4... Phoneme processing unit, 5... Memory, 6...
Interpolation processing unit, 8...Acoustic tube model unit, 9...Calculation unit, IO...Speech synthesis waveform unit, 11...Speaker, P
···sound source.

Claims (1)

[Claims] (1) In a speech synthesis method that uses phonemes as the basic unit and simulates them using sound tubes, one phoneme is divided into a predetermined number of parts, and the input character data is analyzed by referring to the built-in language dictionary. a language processing unit that performs language analysis; syllable processing of data that has syllable parameters and the classified phoneme parameters and is analyzed by the language processing unit;
It has an audio processing unit that performs phoneme processing and a memory that stores information on an exponential function expressed by a predetermined exponential curve, and extracts the information in the memory at an extraction interval corresponding to the processing result of the audio processing unit. and an interpolation processing unit that determines execution parameters by determining the execution parameters, and an interpolation processing unit that determines the execution parameters by determining the execution parameters determined by the interpolation processing unit. 1. A speech synthesis method, comprising: a calculation section that performs calculations at a predetermined timing based on parameters; and a speech synthesis signal is obtained based on the calculation result of the calculation section.