JPS6040636B2 - speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a speech synthesis device, and its purpose is to synthesize a plurality of different pitches in response to each compression parameter without increasing the storage capacity of a data storage unit. An object of the present invention is to provide a speech synthesis device that can selectively synthesize speech.

In general, there is a middle parameter (hereinafter referred to as the A parameter) that represents the magnitude of the sound by sampling the audio signal using a sampling pulse with a frequency higher than the audio frequency, and a pitch parameter (hereinafter referred to as the P parameter) that represents the pitch or fundamental period of the sound. parameter) and a spectral parameter (hereinafter abbreviated as S parameter) representing the timbre of the sound, that is, the spectral distribution,
For playback, each feature parameter is compressed to the number of bits corresponding to the degree to which it contributes to sound quality and stored in the data storage unit as a compression parameter, and each feature parameter is stored in advance as a compression parameter read out sequentially from the data storage unit. In this type of speech synthesizer, which accesses the ROM and synthesizes speech by driving the sound source using characteristic parameters read from the playback ROM, even if the speech differs only in pitch (basic period), As in the case of reproducing different sounds, it is necessary to store compression parameters corresponding to sounds of each pitch in the data storage unit.

Therefore, in order to be able to reproduce sounds at pitches that match the surrounding noise conditions or the user's preferences, it is necessary to store compression parameters in the data storage unit for each pitch of sound. However, there was a drawback that the storage capacity of the data storage unit had to be made larger than necessary. The present invention has been made in view of the above drawbacks. An embodiment of a PARCOR type speech synthesizer will be described below with reference to the drawings. In the PARCOR type speech synthesis method, as shown in FIG. 1, the speech signal Vs is processed by sampling pulses at an appropriate period t. and the sampled value is between Xt and Xt-p (P-
1) Excluding the correlation due to the sampling values, Xt
PARCOR coefficient (partial autocorrelation coefficient: hereinafter abbreviated as K parameter) that extracts only the correlation between and Xt-p
The S-parameter is used to synthesize speech, and the K-parameter is one frame in which the speech can be considered to be in a steady state (
5 to 20 msec) at an appropriate period t. (about 100
Yama sec) Sampling the audio signal Vs on the mirror,
Let K be the correlation coefficient between adjacent sampling values, and between multiple isolated sampling values, the influence of the sampling values sandwiched between them is calculated by linear prediction using the least squares error, and the correlation created by subtracting them is calculated. The relation number is K
2~K,. That is. This K parameter is K,
, K2, K3, which represent the partial autocorrelation with points close to ×t, contain a wealth of information regarding the spectral distribution, but points far from above ×, such as K3, K9, K, o, Since the partial autocorrelation coefficients with points do not contain much information about the spectral distribution, we assign a large number of quantization bits to low-order K parameters and a small number of quantization bits to high-order K parameters. Therefore, it is more effective to reduce the number of bits and reduce redundancy. Therefore, the PARCOR method has a better band compression rate than the autocorrelation coefficient method, which uses an autocorrelation coefficient as an S parameter and allocates the same number of bits to each coefficient.

Typically, each A, P, and K parameter is stored or transmitted in a compressed manner, with 5 bits for the A parameter, 6 bits for the P parameter, and each number of K parameters K, , K2.
・・・・・・7, 6, 5, 4, 4 for K, o
Assign 4, 3, 3, 3, 3 bits, etc. The configuration of the present invention-embodiments will be described below with reference to the illustrated embodiments.

FIG. 3 is a block diagram of a speech synthesis device according to an embodiment. As shown in the figure, this speech synthesis device has a data storage section 8.
It consists of two chips: a control ICA including a control IC and a voice synthesis IC (excluding the dotted lines A and B), and data is exchanged between the two in a bit-serial manner. All audio feature parameters are in playback ROM
It is stored as 10-bit data in I, and the playback ROMI has a memory for corrected pitch in which corrected pitch parameters (hereinafter abbreviated as Pm parameters) for synthesizing corrected speech whose pitch has been corrected are stored. Standard pitch parameters (
A standard pitch storage unit is provided in which the standard pitch storage unit stores the P parameters. The number of data assigned to each feature parameter is optimally distributed according to the degree to which the feature parameter contributes to sound quality. Figure 4 shows the ROMI for playback.
The number of data of each feature parameter Pm, A, P, K, o to K, stored in the table is shown. For example, in the case of the A parameter, 32 data expressed in 10 bits are stored. Therefore, the number of relative address bits required when accessing arbitrary data of the A parameter is 5 bits. This relative address is called a compressed parameter because it represents the characteristic parameter compressed to the minimum necessary size. In contrast, the actual feature parameters stored in the playback ROMI are called playback parameters. As is clear from the above, the number of bits of the reproduction parameter is 10 bits in common for each feature parameter Pm, A, P, K, o to K, but the number of bits of the compression parameter is A, P, Each parameter of K, o to K, is different, and each parameter is 5,
6, 3, 3, 3, 3, 4, 4, 4, 5, 6, 7 bits (
total of 53 bits). However, the relative address for accessing the Pm parameter is the relative address of the P parameter (
compression parameters). In addition, 3 bits, ie, 8 pieces of data, are reserved in the reproduction ROM as a spare area. Since one set (=53 bits) of such compression parameters is extracted every 20 msec (1 frame), which can be considered as an almost steady state of the audio signal, it is possible to record the audio signal at a rate of at most 2650 bits/second, and there is no sound. If sections and beat sections are taken into account, it is actually possible to record audio signals at about 1600 bits/second. Such compression parameters (i.e., relative addresses of the playback ROMI) are read from the data storage section 8 and stored bit-serially in the ring register 3 via the switching circuit 10 for each frame. Since it is not possible to retrieve the stored data from the playback ROMI using only such relative addresses, the first addresses stored in the index ROM 2 as shown in FIG. 5 are sequentially retrieved under the control of the address counter 11. The absolute address (9 bits) of the playback ROMI is calculated by adding the above relative address and the addition circuit 4, and the playback ROMI is added using the absolute address.
I am trying to access it.

By the way, in the embodiment, the basic period generation method is changed when synthesizing standard speech and when synthesizing corrected speech, and when synthesizing corrected speech, the input from the control ICA is changed. A pitch control code is added to the beginning of the compressed A parameter among the compression parameters to be compressed, and a correction signal V is output when the pitch control code is detected.
When M is obtained, the first address of the absolute address is set to 0 from the pitch switching circuit 3 to which this pitch correction signal VM is input.
The adder circuit 4 is controlled so that the Pm parameter is read from the corrected audio storage section of the playback ROMI using the compression parameter of the P parameter. On the other hand, when the correction signal VM is not obtained, the playback ROMI
The P parameter will be read from the standard voice storage section. Here, the Pm parameter is a parameter for making the synthesized corrected voice higher or lower by a certain correction ratio, and in the example, the correction ratio is set to 110% and the corrected voice is corrected to be higher pitched than the standard voice. It looks like this. However, a method for synthesizing speech having a fundamental period corresponding to the P parameter or the Pm parameter will be described later. Note that the correction ratio can be set appropriately, and when setting multiple types of correction ratios (for example, 120%, 10%, 10%, 120%), the capacity of the correction audio storage unit must be multiplied. At the same time, the pitch control code may be made into a plurality of bits so that a P parameter or a plurality of Pm parameters to be read outside the compressed p parameter can be arbitrarily selected. Furthermore, a pitch changeover switch may be provided in place of the pitch control code detection circuit 9. R for reproduction below
The operation of reading playback parameters stored in OMI will be described in detail. The index ROM2 stores the bit allocation number of compression parameters as a 3-bit binary number.
One common bit is provided to reduce the storage capacity of the reproduction ROMI, and a spare bit corresponding to a spare area within the reproduction ROMI is also provided. Data regarding the bit allocation number of the compression parameter is sent to the reproduction control circuit 12, and the reproduction control circuit 12 sends a shift clock to the ring register 3 by the bit allocation number. Therefore, from the ring register 3, depending on the bit allocation number, for example, 5 bits for the A parameter, 6 bits for the P parameter, K, etc. Compressed parameters (relative addresses) are serially sent to the adder circuit in the form of 3 bits...K in the case of parameters and 7 bits in the case of parameters. ring register 3
consists of dynamic shift registers to take up as little chip area as possible. Also index R
R for reproducing each feature parameter stored in OM2
The leading address in the OMI is sequentially sent bit by bit to the addition circuit 4 via the parallel-serial conversion circuit 13, so that an absolute address is calculated by sequentially adding bit by bit. The absolute address consisting of the serial data thus calculated is the serial-to-parallel converter 1.
4, it is converted into parallel data, and the ROMI for reproduction can be accessed. By the way, the regeneration RO
The feature parameters output from M1 are updated every frame, but if the feature parameters change discontinuously at the connection point between each frame when updating data, distortion will occur in the audio signal. Since there is a risk that the clarity may deteriorate, an interpolation calculation circuit 5 is provided to perform approximate linear interpolation at 8 points within 17 frames so that the feature parameters can change smoothly when updating data. There is.

Note that this interpolation calculation circuit 5 does not operate when the corrected speech is synthesized. This interpolation calculation circuit 5 is controlled by a timing control circuit 528, and the timing control circuit 28
As shown in the figure, 8 O-clocks (2.5 seconds) are generated during one frame (20 wsec), and 29 P-clocks (P clocks for parameter input) are generated during one D clock (2.5 seconds).
100 French sec), and 22 in one P clock 0
T-blocks (4.5 rsec) for bit reduction are created. Data is read into the ring register 3 at the first D of the eight D clocks.

Each compression parameter A, P, K,. ......K, is read sequentially at odd-numbered P clocks,
For example, the A parameter is read using five T clocks from T6 to T, o in the P section. Even-numbered P clocks or T clocks other than the above are used by the interpolation calculation circuit 5 and the sound source RO.
This is used as the timing for M6, digital filter 7, etc. 2.5 by the above interpolation calculation circuit 5
Each characteristic parameter updated to a new value every sec is temporarily stored in the P latch 16 and the AK latch 23, respectively. However, all parameters that are not required for the time being for interpolation calculation are transferred to the AK parameter stack 24 and stored as data for speech synthesis in the digital filter 7. On the other hand, the data regarding the fundamental period of the voice stored in the P latch 16, that is, the Pm and P parameters, are preset in the preset type subtraction counter 7. The clock of this subtraction counter 17 is changed by a clock switching circuit 17a to a standard audio clock (of a frequency equal to the sampling pulse).
P clock) and a corrected audio clock (T clock) having a higher frequency than the sampling pulse. controlled by The address counter 18 of the sound source ROM 6 is reset by the 0 output signal VR of the subtraction counter 7, and the sound source control data is transferred from the sound source ROM 6 at a basic cycle corresponding to the period of the 0 output signal VR of the subtraction counter 7. The audio sound source 19 is driven by the sound source control data that is sequentially read out and has the basic period, thereby generating audio sound that has the basic period. The sound source control data is data for faithfully reproducing the tone by reproducing the residual waveform obtained by frequency analysis of the original sound. On the other hand, if the voice does not have a fundamental period, the sound source control circuit 20' drives the iron switching circuit 22 and switches to the silent sound source 21. The unvoiced sound source 21 is white noise that does not have a fundamental period.
is generated. Next, the A parameter and the K parameter are supplied to the digital filter 7, which reproduces the sound by adding information regarding amplitude magnitude and spectral distribution to the signal V supplied from the sound source circuit. In FIG. 3, 25 is an amplifier, 26 is a speaker, and 27 is a crystal oscillation circuit. The operation of the basic period generator for standard speech and corrected speech will be specifically described below.

Now, if the pitch correction signal V is obtained from the pitch control code detection circuit 9, and if not, the P latch 16, which stores data for setting the basic period of the voice, receives data from the standard voice storage section of the playback ROMI. The P parameter (integer) to be read is latched, and the clock of the subtraction counter 17 is switched to the standard audio clock, that is, the P clock (100 msec).

Therefore, the period of the 0 output signal VR of the subtraction counter 17 is an integral multiple of 100 rsec, and the address counter 18, which is reset by this 0 output signal VR, outputs the sound source ROM.
The sound generated based on the sound source control data read from 6 has the above period. For example, P parameter
If evening is ``25'', the fundamental period is 100 x 25 ms.
(Fundamental frequency 400 days 2). On the other hand, when the pitch correction signal VM is obtained from the pitch control code detection circuit 9, P
The latch 16 latches the Pm parameter (integer value) read from the corrected audio storage section of the playback ROMI, and the clock of the subtraction counter 7 is changed by the clock switching circuit 17a to the corrected audio clock, that is, the T clock (4. 5 French seconds). Therefore, the period of the 0 output signal VR of the subtraction counter 17 is an integral multiple of 4.5 usec. In this case, the Pm parameter read from the corrected audio storage section is "61" when the compressed P parameter is read out from the standard audio storage section as P parameter "25".
If the Pm parameter is "61", the 0 output signal V is output from the subtraction counter 17 at a period of 4.5 x 61 French Sec.
R is obtained, and the address counter 18 outputs the sound source RO.
The basic period of the sound generated based on the sound source control data read from M6 is 4.5 x 61 rsec (364Hz).
), and the corrected audio that is corrected to the bass side by about 110% is synthesized. In this case, Pm parameter "61"
corresponds to the P parameter "27.451," and is used to synthesize a voice that is corrected to be about 10% lower than the low-pitched voice.

Incidentally, although the corrected speech synthesized as described above has no problem with respect to the fundamental period, there is a slight problem when spectral information based on the K parameter is added by using the digital filter 7. That is, since the calculation in the digital filter 7 is performed in synchronization with the P clock, if the address counter 18 is reset without synchronization with the P clock, the calculation in the digital filter 7 is performed in synchronization with the P clock.
An error occurs in the calculation process, causing distortion in the synthesized voice. Therefore, in this embodiment, the 0 output signal VR output from the subtraction counter 17 is sent to the address counter 18 via a reset pulse generation circuit 40 as shown in FIG.
I am trying to input it to the reset terminal of. This reset pulse generation circuit 40 is formed by inverters 41a, 41b, a capacitor 42, a NAND gate 43, a D flip-flop 44, and an AND gate 45.
As shown in the time chart of FIG. 1A, the P clock immediately after the 0 output signal VR is obtained from the subtraction counter 17 is output as the reset pulse VR' of the address counter 18. In the figure, A is the 0 detection signal VR when synthesizing standard speech with P parameter "12", and mouth is Pm parameter "2" to correspond to P parameter "12.8".
0 detection signal VR when the corrected voice is synthesized based on ``84'', and C indicates the IJ set pulse VR when the corrected voice is synthesized based on the above.

As described above, since the reset pulse VR' outputted from the reset pulse generation circuit 40 is synchronized with the P clock, the reset intervals of the address counter 18 are not equal intervals, and the 0 detection signal VR The fundamental frequency of is 4
．． In the case of 5 x 248 fsec, the address counter 18 is reset by counting the P clock by one stitch;
The number of cases where 12 P clocks are counted and then reset occurs at a ratio of 4:1.

Therefore, the sound source ROM 6 is addressed at a fundamental period equivalent to the P parameter "12.8", and the voiced sound source 1 is
9 will be controlled, and a corrected speech having a predetermined fundamental period will be synthesized. In addition, the time chart shown in FIG. 7b shows the 0 detection signal VR and the set pulse VR.
' This is to explain the relationship between
The reset pulse VR' corresponding to R is shown.

As is clear from the figure, the 3rd, 6th, 8th, 11th, 1416th... pulse of the P clock is output as the reset pulse VR'. Since the sound source ROM 6 is addressed by the address counter 18 which is reset by this reset pulse VR', the sound source ROM 6 is equivalently 3.7 pitch ratio (
sound. The voiced sound source data will be read out at a period that can be considered as 1), the voiced sound source 19 will be driven at a predetermined fundamental frequency, and the corrected sound will be synthesized at an accurate pitch. The present invention is configured as described above, and the playback ROM
a standard voice storage unit that stores standard pitch parameters for synthesizing a standard voice having a standard pitch; and a corrected voice storage unit that stores corrected pitch parameters for synthesizing a corrected voice whose pitch is corrected. A pitch switching circuit is provided to appropriately switch and control the access method of the playback ROM so that the pitch parameter read from the playback ROM based on the compression parameter becomes the standard or corrected pitch parameter. It is possible to selectively synthesize multiple types of voices with different pitches according to each compression parameter without increasing the capacity, and with a simple configuration, it is possible to synthesize voices with pitches according to the surrounding noise condition or the user's preference. There is an advantage in that it is possible to provide a speech synthesizer that obtains the desired results.

[Brief explanation of the drawing]

FIG. 1 is a diagram explaining the principle of a speech synthesis method according to an embodiment of the present invention.
FIG. 2 is an explanatory diagram of the same operation as above, FIG. 3 is a block circuit diagram of same as above, FIGS. 4 and 5 are diagrams showing the configuration of the playback ROM and index ROM of same as above, and FIG. The circuit diagram and FIGS. 7a and 7b are explanatory diagrams of the same operation. 1 is a reproduction ROM, 8 is a data storage section, 19 and 21 are sound sources, and 30 is a pitch cutting circuit. Figure 1 Figure 2 Figure 7 Cloudiness Figure 4 Figure 6 Figure 5 Figure 7

Claims (1)

[Claims] 1 Sampling the audio signal using a sampling pulse with a frequency higher than the audio frequency to extract feature parameters consisting of amplitude parameters, pitch parameters, and spectral parameters, and assigning the number of bits for each feature parameter according to the degree to which it contributes to sound quality. The compression parameters are stored in the data storage unit as compression parameters compressed into
In a speech synthesis device configured to synthesize speech by agitating a sound source using characteristic parameters read from M, a standard for storing standard pitch parameters for synthesizing standard speech having a standard pitch in the playback ROM. An audio storage unit and a corrected audio storage unit that stores corrected pitch parameters for synthesizing corrected audio whose pitch has been corrected are provided, and the pitch parameters read from the playback ROM based on the compression parameters are standard or corrected. 1. A speech synthesis device comprising a pitch switching circuit that appropriately switches and controls an access method of a playback ROM so as to match a pitch parameter.