JPS5949596B2 - Audio parameter playback control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a reproduction control method for audio parameters in a so-called PARCOR type speech synthesizer.

In general, parameters representing the characteristics of a sound include an amplitude parameter representing the magnitude of the sound, a pitch parameter representing the pitch of the sound, that is, the fundamental period, and a vector parameter representing the timbre of the sound, that is, the spectral distribution.

Conventionally, in order to express the spectral distribution of speech, as shown in Fig. 1, an autocorrelation coefficient Sp between a sample value X of the speech signal and a sample value A correlation coefficient method has been developed. However, since the autocorrelation coefficient Sp also includes correlations due to (p-1) sample values between Xt(5xt-p), it has the disadvantage of high redundancy and poor band compression rate Therefore, PARCO extracts only the correlation between Xt and X, -p, excluding the correlation between (p-1) sample values between Xt and
R coefficient (partial autocorrelation coefficient) Kp. To explain in more detail, the PARCOR coefficient is defined as ``In one frame (approximately 20 mS) where the audio can be considered to be in an almost steady state,
The audio signal is sampled at certain time intervals (approximately 100 μS), and the correlation coefficient between adjacent sample values is expressed as K,
Between sample values separated by a plurality of intervals, the influence of the sample values sandwiched between them is determined by least squares prediction, and the correlation coefficients K2 to KIO are obtained by subtracting them. Note that in a normal PAROCR type speech synthesizer, the amplitude parameter, pitch parameter, and PARCOR coefficient are abbreviated as A parameter, P parameter, and K parameter, respectively, so these abbreviations will be used in the following description of the present invention. Make it. By the way, among the PARCOR coefficients Kp, K,,K2,K. The coefficients representing the partial autocorrelation with points close to Xt, such as The relationship number does not contain much information about the spectral distribution. Therefore, conventionally, a large number of quantization bits are assigned to low-order K parameters such as Kl, K2, and K3, and a small number of quantization bits are assigned to high-order parameters such as K8, K9, K, and O. It is known that after K, the spectral distribution can be reproduced with sufficient accuracy without transmitting any more. Therefore, PARCO
The R method has a better band compression rate than the autocorrelation coefficient method which requires the same number of bits for each of the coefficients S1 to SIO. The present invention and its combined inventions provide an extremely simple playback control method in which feature parameters are recorded in a linearly or non-linearly compressed form and the synthesizer reproduces the original feature parameters from the compressed parameters in such a PARCOR type speech synthesizer. The first purpose is to provide a system for applying a PARCOR type speech synthesizer using such a simple playback control method to a time signal device, an alarm device, an alarm clock, etc. The LSI for speech synthesis is made into a two-chip structure consisting of an LSI for the prongs including a data recording section and an LSI for speech synthesis.
When providing versatility to the SI, it is possible to arbitrarily select the sound quality and length of the recorded and reproduced metsuji simply by changing the ROM section, and also to transfer data between the LSI for the torpedo and the LSI for speech synthesis. The second objective is to reduce the number of pins in the LSI package, simplify assembly, and reduce the chip area by performing internal LSI calculations bit-serially. Redundancy is further reduced by subdividing frequently-frequent parts of low-order K parameters that have an impact and performing non-linear compression, and the compressed parameters are made to correspond one-to-one to the original feature parameters. A third purpose is to reduce the storage capacity of the reproduction ROM by sharing data in the reproduction ROM for reproduction between different compression parameters. DESCRIPTION OF THE PREFERRED EMBODIMENTS The configuration of the present invention and its combined invention will be described below with reference to illustrated embodiments.

FIG. 3 is a block diagram of a speech synthesis apparatus using the speech parameter regeneration rambling method according to the present invention and its combined invention. As shown in the figure, this speech synthesis device is composed of two chips: an FbI leg 1C (A) including a data recording section 40, and a speech synthesis 1C (excluding dotted line parts A and B). Data is exchanged between the two in bit serial format. All voice characteristic parameters are stored as 10-bit data in the playback ROM, and the number of data assigned to each characteristic parameter is optimally distributed according to the degree to which that characteristic parameter contributes to sound quality. . FIG. 4 shows the number of data of each characteristic parameter of A, P, and KlO-Kl stored in the reproduction ROMl. For example, in the case of A parameter 1
Thirty-two pieces of data expressed by 0 bits are recorded.
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. On the other hand, the actual characteristic parameters stored in the playback ROM1 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 of the feature parameters A, P, KlO-K, but the number of bits of the compression parameter is A, P, KlO-K. , are different for each parameter of 5, 6, 3, 3, 3, respectively.
3, 4, 4, 4, 5, 6, 7 bits (total 53 bits)
It is. In addition, a reserve area of 3 bits, ie, 8 pieces of data, is reserved in the reproduction ROM. Since one set (=53 bits) of such compression parameters is extracted every 20 msec (one 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, resulting in no sound. If sections and repeat 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 ROM1) are stored in series in the ring register 3 from the data input terminal 8 via the switching circuit 10 for each frame, but only such relative addresses R for playback with
Since stored data cannot be retrieved from OMl,
The leading addresses stored in the index ROM 2 as shown in FIG. An address (9 bits) is calculated, and the reproduction ROM 1 is accessed using the absolute address.

As shown in FIG. 5, the index ROM 2 stores the number of bits allocated for compression parameters as a 3-bit binary number, and is also used to reduce the storage capacity of the playback ROM 1, which is the main part of the invention merged with the present invention. One common bit is provided, and a spare bit corresponding to a spare area in the reproduction ROM1 is also provided. Data regarding the bit allocation number of the compression parameter is sent to the reproduction leg circuit 12, and the reproduction control circuit 2 sends shift locks to the ring register 3 by the bit allocation number. Therefore, from the ring register 3, depending on the above bit allocation number, for example, in the case of the A parameter, 5 bits, and in the case of the P parameter, 6 bits are sent.
In the case of bit, K, O parameters, 3 bits...
In the case of K and parameters, compressed parameters (relative addresses) of 7 bits are each sent serially to the adder circuit. The ring register 3 is composed of a dynamic shift register so as to occupy as little chip area as possible. Furthermore, the leading address in the reproduction ROMI of each characteristic parameter stored in the index ROM 2 is determined by the parallel-serial conversion circuit 13.
Since the bits are sequentially sent to the adder circuit 4 via , the absolute address is calculated by sequentially adding bits one by one. The serial absolute address thus calculated is converted into parallel data via the serial/parallel converter 14, so that the reproduction ROMI can be accessed. By the way, regarding how to arrange the 10-bit playback parameters in the playback ROMI, in the case of A parameters, P parameters, and high-order K parameters such as KIO to K1, it is okay to arrange the data at almost equal intervals. do not have. However, for low-order K parameters such as K,, K2, and K3, it is inconvenient to arrange the data at equal intervals. This is because, as shown in Figure 6, the audio frequency normally used by humans is the sampling frequency (e.g. 8KH).
z), the K1 parameter, which represents the correlation between adjacent sample values (Xt and Xt-1), takes a value almost equal to 1, minimizing the influence of the K, The K2 parameter, which represents the correlation between Xt and X,-2 when removed by the linear prediction method using multiplicative error, has a value almost equal to -1 because the slope from Xt to X,-2 hardly changes. This is because it has the characteristic of taking . Therefore, for example, in the case of the K1 parameter, many values close to 1 are converted into 10-bit digital data and stored in the playback ROMI, while values from -1 to o are not often stored. Similarly, regarding the K2 parameter, a large amount of data for values close to -1 is stored in the reproduction ROMI, and not much data is stored for values from o to 1. In the present invention, each of the parameters K, K2, and K3 is subdivided according to the number of frequencies, that is, nonlinear compression is applied, but applying such nonlinear compression itself is already known. By the way, when each compression parameter corresponds to a playback parameter on a one-to-one basis in this way, a considerably large storage capacity is required in the playback ROMI.

Of course, there is no particular problem with high-order K parameters such as K8, K8, and KIO, since the number of data is only 8 each, but low-order K parameters such as K, and K2 each have the number of data. There are 128 and 64 pieces,
This requires an incomparably large storage capacity compared to high-order K parameters. Therefore, by taking advantage of the fact that the frequency distribution of the K1 parameter is very similar to the frequency distribution of the K2 parameter if the positive and negative signs are reversed, the data in the ROMI for reproducing the K parameter can be shared as the data for the K2 parameter. However, the purpose of the present invention and its combined inventions is to reduce the storage capacity of the playback ROMI. To specifically accomplish this, a circuit configuration as shown in FIG. T is used.

This circuit operates in the same way as the conventional example when the common bit 0D provided in the index ROM 2 is o, but when the common bit 0D is 1 as shown at K2 in FIG. behaves differently.
First, the compression parameters sent bit-serially from the ring register 3 are inverted to logical values 1 and 0 by the bit inversion circuit 29. Figure 8 shows bit inversion circuit 2.
As shown in the figure, when the common bit 0D is 1, the logical value of the compression parameter is inverted and output. Further, a request signal (shift clock) for extracting compression parameters bit by bit from the ring register 3 is delayed by a timing corresponding to one bit by a one-bit delay circuit 30. However, since the absolute address sent bit by bit from index ROM2 is sent without delay, the K2 parameter, which is originally 6 bits, is expanded to 7 bits and added to the absolute address. . Moreover, as shown in Figure 5, the starting address of the K2 parameter is K
, is set to the same address as the start address of the parameter, so the relative address of the K2 parameter is carried up by one bit, which is less than the K1 parameter, and the data related to the K parameter in the playback ROMI is stored. Every other piece of data will be accessed. The reproduced value of the K2 parameter thus reproduced by using the data of the K1 parameter is sent to the interpolation calculation circuit 5 after its sign is inverted by the positive/negative inversion circuit 31. FIGS. 9A and 9B are timing charts showing these series of operations. As shown in FIG. 9A, when the LOAD signal is input, the request clock signal CLr8:1 is sent to the reproduction control circuit 12; The up counter in (1
11), that is, the request signal is output in an amount corresponding to the number of bits allocated to each compression parameter. However, when the common bit 0D is 1, the NAND circuit 32 inhibits the inputs of the AND gates 33 and 34. The first request signal is not outputted, and only after one request clock signal CLreq is input and the flip-flop 35 is inverted, the request signal is outputted. The subsequent operation is the same as in the case of FIG. 9a, and the request signal is output until the up counter in the reproduction Fbl leg circuit 12 reaches (111), resulting in a timing delay of 1 bit. A request signal will be output. In the present invention, whether or not to perform such a series of operations is determined based on the common bit 0D stored in the index ROM 2.
By storing 1 in the K2 parameter portion of the common bits in the index ROM 2 and setting the start address to the same address as the K1 parameter, the data related to the K2 parameter in the playback ROM 1 ( 64) can be reduced. By the way, the feature parameters output from the playback ROM1 are updated every frame, but if the feature parameters change discontinuously at the connection points between each frame when updating the data, distortion may occur in the audio signal. Therefore, an interpolation calculation circuit 5 is provided to perform approximate linear interpolation at 8 points within one frame so that the feature parameters 5 can change smoothly when updating data. That's what I do.

Therefore, the timing leg circuit 28 generates 8 interpolation D clocks 5 (2.5 msec) during one frame (20 msec), and 25 interpolation clocks 5 during one D clock, as shown in FIG. P clock for reading parameters (100
Further, 22 bit reading T clocks (4.5 μSec) are created in one P clock.
Data is read into the ring register 3 from the data input terminal 8 at the first D1 of the eight D clocks.

Each compression parameter A, P, K, O...K1 is read sequentially at the odd-numbered P clock.
The parameters are read at five T clocks from T5 to T and O in the P1 section. The even-numbered P clocks or T clocks other than those mentioned above are used as timing for the interpolation calculation circuit 5, the sound source ROM 6, the digital filter 7, etc. 2.5 msec by the interpolation calculation circuit 5
Each characteristic parameter updated to a new value 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 calculations 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 is transferred to the sound source ROM 6 via a matching circuit 17 and a counter 18.
and if the voice has a fundamental period, it is sent to the voiced sound source 19.
is driven to generate a pulse signal having the above basic period. If the sound has no fundamental period, the sound source control circuit 20
The switching circuit 22 is driven to switch to the silent sound source 21. The unvoiced sound source 21 is white noise (
white noise). 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 supplied from the sound source circuit. In addition, in Fig. 3, 25 is an amplifier,
26 is a speaker, 27 is a crystal oscillation circuit, and 9 is a parameter code detection circuit, but since these are not directly related to the gist of the present invention, detailed explanation thereof will be omitted. The present invention is configured as described above, and when compressing audio waveform information and recording it in the data recording section, each compression parameter is configured as a relative address of a reproduction ROM, and each characteristic parameter group is reproduced. Index RO that stores the start address in the ROM in the same order as the data recording section
By sequentially incrementing the address counter of this index ROM, the corresponding start address is added to each compression parameter (relative address) read as data, and this is added to the playback ROM.
Since the original feature parameters are played back as the absolute addresses of It has the advantage of not requiring complicated calculations, and in particular, the value of each feature parameter to be stored in the playback ROM can be arbitrarily selected, so it has the advantage of being able to finely divide parts with high frequency of each parameter and perform non-linear compression. be.

In addition, the bit allocation number of the compression parameter is recorded in binary in one word of the index ROM, so when reading data, by simply incrementing the address counter, the bit allocation number is decoded and the request pulse required for data reading is generated. It has the advantage that it can output the required number of shift clocks when serially adding an absolute address and a relative address with a 1-bit adder. In addition, in the combined invention related to the improvement of the present invention, a playback ROM that records playback parameters is provided.
A spare area is provided in the index ROM, and a spare word is provided in the index ROM that is added to the compression parameters to provide an absolute address for accessing the playback ROM.Therefore, there is no need to change the circuit depending on the application, and for example, it is possible to write short messages. If clarity is important, reduce the number of spare bits.
, and the bit allocation of the K2 parameter one bit at a time, or change the bit allocation of each K parameter depending on individual differences in sound quality. LS for synthesis
It has the advantage of being able to mass produce I.

In addition, one word of data recorded in the index ROM contains the start address of each parameter as well as the binary value of each compression bit number, so it is possible to create shift clocks for the number of compression bits at the same time as reading the start address. This has the advantage that it is easy to configure a circuit that reads the necessary number of unevenly distributed compression parameters bit-serially. It has the advantage that a two-chip system with various types of portable LSIs, such as those for commercial use, intercom use, and alarm use, can be constructed at low cost with an extremely small number of connection pins. Further, in another merged invention related to an improvement of the present invention, the number of data of K, which stores data related to parameters, and a ROM for reproducing parameters is allocated to a large number in the range of 0 to 1, and in the range of -1 to o. Allocate less and use the RO for regenerating the K1 parameters.
The address signal of M is set as the compression parameter of the K parameter, and by accessing the ROM using the compression parameter, data regarding the K parameter is reproduced, and the number of bits of the compression parameter of the K2 parameter is set to K.
It is smaller than in the case of 1 parameter, and the negative logic of the compression parameter of the K2 parameter is used, and it is carried by the number of bits that is insufficient than the compression parameter of the K1 parameter, and the K1 parameter is reproduced using the carried data as an address. The K2 parameter reproducing ROM is
can be omitted, and data related to K2 parameters (which previously occupied a large storage capacity in playback ROMs) can be omitted.
64) can be reduced.

In other words, the present invention focuses on the fact that the voice frequencies normally used by humans are largely distributed in a frequency range sufficiently lower than the sampling frequency, and based on this, the frequency distribution of the K1 parameter and the K2 parameter is physically determined. Taking into account the fact that their frequency distributions are always complementary, data in the playback ROM that matches the frequency distribution of K and parameters is transferred to K via a predetermined inverting circuit.
By sharing the two parameters, the storage capacity of the playback ROM is greatly reduced, and the LS for speech synthesis
This greatly contributes to the ease of manufacturing I.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram of the principle of PARCOR coefficients used in the present invention, FIG. 2 is an explanatory diagram of each clock used in the above, FIG. 3 is an overall block diagram of an embodiment of the present invention and its combined invention, and FIG. Figure 5 is an explanatory diagram showing the configuration of the reproducing ROM used in the above, Figure 5 is an explanatory diagram showing the configuration of the index ROM used in the same, Figure 6 is a graph showing the frequency distribution of PARCOR coefficients used in the same, and Figure T 8 is a partially enlarged circuit diagram of the same, and FIGS. 9A and 9B are time charts of the same. 1 is a playback ROM) 2 is an index ROM) 3 is a ring register, 4 is an addition circuit, 5 is an interpolation calculation circuit, 6 is a sound source ROM, . T is a digital filter.

Claims (1)

[Scope of Claims] 1. Each characteristic parameter representing the amplitude, fundamental period, spectral distribution, etc. of a sound waveform is compressed to the number of bits corresponding to the degree of contribution to sound quality and recorded in a data recording section, and from the data recording section. In an audio parameter reproduction control method that reproduces audio by reproducing original characteristic parameters from sequentially read compression parameters, a reproduction ROM in which a predetermined number of values of each characteristic parameter is stored in advance.
Then, an index ROM is provided in which the leading address of each feature parameter group of this playback ROM is stored in the same order as each compression parameter arranged in the data, and the relative address in the playback ROM of each feature parameter is stored. The compression parameters correspond to the characteristic parameters, and each compression parameter in the above data read from the data recording section is added to the top address sequentially read from the index ROM, thereby corresponding to the compression parameters in each data from the playback ROM. 1. A sound parameter playback control method, characterized in that an absolute address for reading out characteristic parameters is created. 2. Each audio parameter representing the amplitude, fundamental period, and spectral distribution of the audio waveform is quantized with the same number of bits, and each audio parameter is compressed linearly or nonlinearly with non-uniform bit allocation, and each audio parameter value is quantized one-to-one. In the audio parameter playback control method, the audio waveform data is read in the form of the compression parameter and reproduced to the original audio parameter to reproduce the audio waveform from each audio parameter. a playback ROM storing audio parameter values of the compressed number of bits each time, and one word containing the start address of each audio parameter group in the playback ROM and a binary value of the compressed bit number. and an index ROM having a number of words equal to the number of types of audio parameters plus one word for a spare bit, and the compression parameter is defined as the amount of displacement of the corresponding audio parameter from the top address in the playback ROM, and An address counter is provided to sequentially read each word from the index ROM, and the same number of data request pulses as the number of compressed bits read out is generated to read audio waveform data consisting of compressed parameters bit serially. The playback ROM is accessed using an address formed by adding the first address read from the index ROM to the compression parameter to play back each audio parameter corresponding to the read data, and the data is read from the index ROM. A reproduction control method for audio parameters, characterized in that the data request pulse is not generated when a word for a bit is read out, and a spare area corresponding to the spare bit is provided in a playback ROM. . 3 Compression parameters representing the amplitude, fundamental period, spectral distribution, etc. of the audio waveform are compressed to the number of bits corresponding to the degree of contribution to sound quality and recorded in the data recording section, and are sequentially read out from the data recording section. In an audio parameter playback control method that reproduces audio by reproducing original feature parameters from The partial autocorrelation coefficient between sample values is a first-order coefficient, and the partial autocorrelation coefficient between adjacent sample values separated by one sample value is a second-order coefficient, and 1
ROM for primary coefficient reproduction that can store data related to secondary coefficients
A large amount of data is allocated in the 0 to 1 area, and a small amount is allocated in the -1 to 0 area, and the address signal of the primary coefficient reproduction ROM is used as the compression parameter of the primary coefficient, and the compression parameter By accessing the ROM, the data regarding the primary coefficient is reproduced, the number of bits of the compression parameter of the secondary coefficient is made smaller than that of the primary coefficient, and the negative logic of the compression parameter of the secondary coefficient is applied. , carry up by the number of bits missing from the primary compression parameter, access the primary coefficient regeneration ROM using the carried data as an address, and invert the sign of the output to obtain data related to the secondary coefficient. A method for controlling reproduction of audio parameters, characterized in that a ROM for reproducing secondary coefficients is omitted while reproducing data.