JPH11296196A - Sound encoding method and sound encoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce power consumption at the time of sound encoding processing by reducing the operation quantity of sound encoding processing, when the sound is encoded by using a code note consisting of the correspondence tables of sound source parameters and codes. SOLUTION: When sound encoding is started, a sound parameter 101d is read out from a code note 101, the sound is reproduced by using a synthesis filter 300, an error from the inputted sound is evaluated by an error evaluation device 301, the evaluated value is compared with a threshold value ε, when the former is smaller than the latter, and a code 101c corresponding to sound source parameters at the time is outputted to the outside. When the evaluated value is larger than the threshold value ε, it is stored in a register 102b, and updated to the minimum successively or all are stored.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice coding method and a voice coding apparatus used in a digital portable telephone, and more particularly to an improvement in coding speed and a reduction in power consumption during coding processing. The present invention relates to an audio encoding method and an audio encoding processing device capable of performing the following.

[0002]

2. Description of the Related Art In recent years, voice coding techniques for compressing voice bandwidth have been used in digital automobile telephones and mobile telephones in which the number of subscribers has increased remarkably. 2. Description of the Related Art Conventionally, as a speech encoding method used for a digital car phone / cellular phone, for example, “Acoustic Society of Japan, Vol.
5), pp. 784-789 ”. In this speech coding apparatus, a codebook (R) storing a large number of sets of speech source parameters and codes corresponding thereto is stored.
OM), the audio parameters are sequentially read from the start address, and the audio is reproduced through the synthesis filter.
In this method, an error between the synthesized voice and the input voice is evaluated by an error evaluation device, and a code corresponding to the sound source parameter determined to have the minimum error is output to a human unit. In this method, the minimum error cannot be determined unless all the sound source parameters have been read from the codebook (ROM).

[0003]

SUMMARY OF THE INVENTION As a basis of various systems of voice coding for car phones and mobile phones, CE is used.
LP (Code Excited Linear Pr)
Edition) system. In the CELP method, speech is encoded in the form of associating various signal sequences stored as a codebook with input speech. This association process is performed by synthesizing voices for all signal sequences in the codebook in accordance with an analysis method based on synthesis, and selecting the best audible sequence. That is, as in the above-described speech coding apparatus, it is necessary to read out all the sound source parameters from the codebook (ROM). Therefore, the amount of calculation becomes enormous in order to perform coding with good sound quality. There is a problem that a high-speed signal processing device is required, and power consumption for processing is increased. Therefore, an object of the present invention is to solve such a conventional problem, reduce the amount of calculation of the audio encoding process, and reduce the power consumption during the audio encoding process. An object of the present invention is to provide a speech encoding processing device.

[0004]

In order to achieve the above object, in the speech encoding method of the present invention, (1) a sound source parameter for synthesizing speech and a plurality of sets of codes corresponding to the sound source parameter are stored. By providing a codebook to perform, by comparing the voice generated from the sound source parameters read from the codebook and the input voice, by evaluating the error,
In a speech coding method for selecting a code to be output based on the evaluation result, when performing a speech coding process using the codebook, the evaluation result of the error is compared with a constant to determine the magnitude relationship. The code is output according to the determination result, or the codebook is divided, and the code is output by reading out the excitation parameters only from some of the divided codebooks, or the address read from the codebook is separated. Or output the code corresponding to the sound source parameter read out from the address, or use one of them, so that the entire codebook is not accessed and the optimum It is characterized in that the speech encoding process is completed by outputting a suitable code (see each embodiment of FIGS. 3, 4 and 8). Also, (2) determining the magnitude relation between the evaluation result of the error and the constant, holding the determination results for the number of codebook entries, and determining the code corresponding to the minimum value of the plurality of determination results. Output or hold the judgment result for the planned capacity equal to or less than the number of entries in the codebook, and after exceeding the planned capacity, update the judgment result to the smallest one, and finally set the minimum value to (See the first embodiment in FIG. 3). Further, in the speech encoding processing device of the present invention,
(3) The codebook is provided in a hierarchical manner, and the codebooks belonging to the upper layer have a smaller capacity than the codebooks belonging to the lower layer. (See the second embodiment in FIG. 4). Further, a storage device for holding the evaluation result of the error, a device for determining a maximum value and a minimum value of data stored in the storage device, a codebook belonging to an upper layer and a lower layer (See the second embodiment in FIG. 4). Also, (4) when performing the exchange of data between the codebook in the upper layer and the codebook in the lower layer, based on the previous speech encoding processing result,
It is characterized in that data determined to have the largest error in the upper layer codebook is replaced with data in the lower layer codebook (see the second embodiment in FIG. 4).

(5) When exchanging data between a codebook in an upper layer and a codebook in a lower layer, a codebook of an upper layer is determined based on the history of the speech encoding processing results so far. The data determined to have the largest error among them may be replaced with the data in the lower-order codebook (see the second embodiment in FIG. 4). (6) At the end of the audio encoding process, the contents of the upper-layer and lower-layer codebooks are saved to another storage device, and at the start of the next audio encoding process, the contents of the storage device are saved to the upper-layer and the lower layer. It is also characterized in that it is configured to be read out to a lower layer codebook (see the second embodiment in FIG. 4). (7) When the contents of the codebooks of the upper layer and the lower layer are saved in another storage device at the end of the audio encoding process, the contents can be stored for each user of the audio encoding device. It is also characterized in that a personal codebook is configured (see the third embodiment in FIG. 6). (8) The personal codebook is constituted by a nonvolatile semiconductor memory device (see the third embodiment in FIG. 6). (9) A storage device for holding the error evaluation result, a device for determining the maximum value and the minimum value among the data stored in the storage device, and a code book data storage address. It is also characterized by having a storage device for holding the unit and a management device for managing addresses registered in the address storage device (see the fourth embodiment in FIG. 8). Further, (10) the storage device for storing a part of the data storage address of the codebook is mainly characterized by storing an address corresponding to a code having a small error evaluation result value (see FIG. 8). Refer to the fourth embodiment).

[0006]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram of a speech encoding processing device showing a basic configuration of the present invention. In FIG. 1, reference numeral 100 denotes an audio encoding processing device for converting an audio signal into a code, and a synthesis filter for reproducing audio from a sound source parameter read from a codebook and a reproduced audio And an error evaluation device for evaluating an error between the input and the input voice. Reference numeral 101 denotes a codebook having the same configuration as that of the related art or newly improved according to the present invention, in which a large number of sets of sound source parameters and codes corresponding thereto are registered. According to the present invention, there is a configuration in which an upper hierarchy and a lower hierarchy are separated. Reference numeral 102 denotes a device newly provided according to the present invention, which includes an error determination device and a register, or a minimum error determination device and a register and a primary / secondary codebook management device, or a personal codebook in addition thereto. It has an address table and the like, and is for controlling the codebook 101.

FIG. 2 is an operation flowchart of a speech encoding method according to an embodiment of the present invention. When instructed to start speech encoding (step 10a), speech encoding processing apparatus 100 accepts speech input (step 11a), and at the same time, reads a code and a sound source parameter set (hereinafter referred to as an entry) from a codebook (step S11a). Step 12a). A speech is synthesized using the read sound source parameters, and an error from the input speech is evaluated by an error evaluation device of the speech encoding processing device 100 (step 13a). As a result, if the obtained value is larger than a predetermined threshold value (step 14a), the next entry is read from the codebook (step 12a), and speech is synthesized again using the sound source parameters. , Evaluate the error with the input voice (step 13)
a). If the error evaluation result is larger than the threshold value, the above processing is repeated, and if smaller than the threshold value, the codebook is controlled and the code corresponding to the excitation parameter at that time is output to the outside (step 15a). . Until the end of the voice encoding is instructed, the processing from the voice input to this point is repeated (step 16a). The difference between the present embodiment and the conventional method is a process of determining whether or not the error is equal to or less than a threshold value. This is the point at which reading from. Compared to reading all the excitation parameters of the codebook conventionally, the processing can be terminated in the middle in the present embodiment, so that high-speed encoding can be performed and power consumption during processing can be reduced. .

FIG. 3 is a block diagram of a speech coding processing apparatus according to a first embodiment of the present invention. In FIG.
Is a codebook, 101c is a code held in the codebook 101, 101d is a sound source parameter associated with the code, and these two constitute one set to form one entry. Reference numeral 300 denotes a synthesis filter that synthesizes speech based on the sound source parameters read from the codebook 101, and 3
Reference numeral 01 denotes an error evaluation device for evaluating an error between the voice synthesized by the synthesis filter 300 and a voice input from the outside. 302 is a device for reading entries from the codebook 101, and 303 is a device for selecting the code 101c output from the codebook 101 to the outside. Reference numeral 102 denotes a codebook control device newly provided in the present invention, which is an error determination device 102a for performing a magnitude determination between the threshold value ε and an error evaluation result output from the error evaluation device 301, and an error determination device 102
a register 102b for holding the determination result of a, and a selection device 102c for selecting and outputting either the output of the error determination device 102a or the output of the register 102b. The code control device 102 controls the device 302 for reading entries from the codebook 101 each time the comparison is completed, and controls the device 303 for selecting the code 101c output from the codebook 101 to the outside based on the result of the comparison. . Here, the register 102b is set to have a capacity capable of storing all the sets of the codebook, and the error difference determination device 102a compares the error output with the threshold ε. If all error outputs are greater than the threshold value ε, the first method of controlling the selection device 303 to output the code corresponding to the smallest one of them is stored in the register 102b. Is made small enough to store all the sets of
When the error is the minimum value among the values stored in the register 102b, there is a second method of sequentially updating the error with the minimum value.

Further, the operation of FIG. 3 will be described in detail. When the audio encoding process is started, first, the first entry of the codebook 101 is read. The synthesis filter 300 synthesizes speech from the read sound source parameters of the entry, and evaluates an error with the speech input by the error evaluation device 301.
At that time, the average error power of each voice may be calculated,
The average error may be calculated after frequency conversion of each voice. Further, the ratio of the average error power to the peak power may be used as an evaluation index. The following processing can be identified based on the magnitude relationship between the value obtained as a result of the error evaluation and a certain threshold value ε. That is, when the error evaluation value is larger than the threshold value ε, the next entry in the codebook is read, and the speech synthesis and the error evaluation process are repeatedly performed. If the error evaluation value is smaller than the threshold value ε, the above processing is terminated, and the code 101c at that time is selected as a code obtained by encoding the input voice, and is output to the outside. At this time, if the error evaluation value does not become smaller than the threshold value ε even if the processing is performed over the entire codebook 101, the error evaluation value corresponds to the one closest to the threshold value ε. The code is selected as a coded version of the input speech and output to the outside. The threshold value ε is a value to be prepared according to the error evaluation method in the error evaluation device 301, and may be a fixed value or a value input as a signal from outside. With the above-described configuration, the speech encoding processing device using the present invention can perform the encoding process without accessing the entire codebook, when considered on average.
That is, it is possible to greatly reduce the amount of calculation and the power consumption during the audio encoding process.

FIG. 4 is a block diagram of a speech coding processing apparatus according to a second embodiment of the present invention. In FIG.
a is a primary codebook holding a part of entries, and 101b is a secondary codebook holding entries other than those stored in the primary codebook 101a. 302 is a primary codebook 1
A device for reading entries from 01a, 303 is a device for selecting a code output from the primary codebook 101a to the outside. Codebook control device 400 includes a register 400a that holds an error evaluation result output from error evaluation device 301,
A minimum error determination device 400b for determining the smallest one among the values in the register 400a, a primary / secondary code for managing entries held in the primary codebook and the secondary codebook, and controlling the replacement thereof The book management device 400c is provided. The codebook control device 400 includes a minimum error determination device 400
Each time b determines the minimum value, it controls the selecting device 303 of the code output from the codebook 101 to the outside, and the primary / secondary codebook management device 400c controls the entry of the primary codebook and the secondary codebook. Entry swapping device 4 for swapping
01 is controlled. Reference numeral 300 denotes a synthesis filter for synthesizing speech based on the sound source parameters read from the primary codebook 101a. Reference numeral 301 denotes an error evaluation device for evaluating an error between the speech synthesized by the synthesis filter 300 and speech input from the outside. is there.

[0011] The primary codebook 101a is a secondary codebook 101b.
The primary codebook 101a and the secondary codebook 101b can exchange entries via the entry exchange device 401. The synthesis filter 300 includes the primary codebook 101a
Are transmitted in order from the first address to the last address, and the error evaluation between the synthesized voice and the input voice is performed by the error evaluator 301, and the error evaluation result is sequentially transmitted to the codebook controller 400. Sent to In the codebook control device 400, the output signal from the error evaluation device 301 is stored in the register 400a, and the minimum error determination device 400b
Determines the entry corresponding to the minimum value among them. The signal output from the minimum error determination device 400b is sent to the code selection device 303, where the corresponding code is selected and output to the outside. In the present embodiment, since the encoding process is completed only by reading the entry of the primary codebook 101a by dividing into the primary codebook 101a and the secondary codebook 101b, the encoding is performed in the same manner as in the first embodiment. The effect of speeding up the conversion process and reducing power consumption can be obtained.

FIG. 5 is a flowchart of the operation of the codebook control device and the codebook exchanging operation in FIG. When the speech encoding process is started (step 10b), first, an entry is read from the head address of the primary codebook 101a (step 11b). Next, speech is synthesized according to a predetermined procedure based on the sound source parameters in the entry. An error between the synthesized voice and the voice input from the outside is determined, and the evaluation result is input to the register 400a and stored (step 12b). Next, an entry is read from the next address of the primary codebook 101a, and the same operation as when reading from the first address is performed. Primary codebook 101a
The above operation is repeated until the end of is reached (step 13b). By doing so, the primary codebook 1
After the processing of the entry at the end of 01a is completed, the register 400a stores the evaluation results of all the excitation parameters of the primary codebook 101a. From the series of evaluation results, the code corresponding to the sound source parameter determined to have the smallest error in the minimum error determination device 400b is selected as a coded version of the input voice, and output to the outside ( Step 14b).
Here, the determination result in the minimum error determination device 400b is also sent to the primary / secondary codebook management device 400c. The primary / secondary codebook management device 400c processes the determination result according to a predetermined procedure, and issues a control signal to the entry replacement device 401 based on the result. In accordance with a control signal from the primary / secondary codebook management device 400c, the entry replacement device 401 changes the entry of the primary codebook 101a with the maximum error and the secondary codebook 101.
Replace with the one on b (step 15b). In addition,
It is assumed that the excitation parameter having the smallest error in the secondary codebook 101b is determined in advance by the management device 400c. Thus, one speech encoding process is completed (step 16b), and the process proceeds to the next speech encoding process.

In FIG. 5, the procedure performed in the primary / secondary codebook management device 400c may be as follows. That is, a flag is added to an entry that has been found to have the largest error from the result of the determination by the minimum error determination device 400b, and after a number of speech encoding processes, the number of flags has reached the specified number. The object is replaced with an entry on the secondary codebook 101b. Entries on the secondary codebook 101b also have similar flags, and the entry with the least number of flags is to be replaced. With the above-described configuration, the codebook used during the speech encoding process is limited to only the small-capacity primary codebook 101a, so that not only high-speed processing but also low power consumption can be achieved. Further, in order to update and optimize the contents of the primary codebook 101a during operation so that the input speech can be encoded with a smaller error, the primary codebook 101a and the secondary codebook 101b have a small capacity. To form the entire codebook, the contents of the conventional codebook not using the present invention and the contents of the codebook can be completely matched, and no inconvenience is caused by using the present invention.

FIG. 6 is a block diagram of a speech coding processing apparatus according to a third embodiment of the present invention. The feature of this embodiment is that a personal codebook 60 for storing the state of the codebook for each user is provided.
1 is provided. A codebook control device 600 has almost the same function as the codebook control device 400 in FIG. 4 and further has a function of controlling the personal codebook 601 according to a personal identification signal input from the outside. . 6
01 is the primary codebook 101a, the secondary codebook 101b, and the codebook control device 6 at the end of the speech encoding process of each speaker.
This is a personal codebook that holds the state of 00. Personal codebook 6
Reference numeral 01 is a large-capacity storage device for holding the states of the primary codebook 101a, the secondary codebook 101b, and the codebook control device 600 for each of a plurality of speakers. When the speaker is determined, a personal identification signal for identifying the speaker is input to codebook control device 600, and a codebook selection signal is generated based on the signal, and the signal is stored in personal codebook 601. By being input, the previous primary codebook 101a and secondary codebook 1
Based on the state of 01b, the entry exchange device 401 performs exchange between the primary / secondary codebooks. As a result, encoding suitable for the user is performed.

FIG. 7 is a flowchart showing the operation of the speech encoding process in FIG. Prior to the start of the voice encoding process (step 11c), a personal identification signal for specifying a speaker is input to the codebook control device 600 (step 12c). The codebook control device 600 generates a codebook selection signal from the input personal identification signal, and selects a corresponding part of the personal codebook 601. The selected part is
The primary codebook 101a, the secondary codebook 101b, and the codebook control device 600 are loaded at predetermined positions (step 13c). Thereafter, speech coding is performed according to the procedure shown in FIG. 5 (step 14c). That is, like the codebook control device 400 of FIG. 4, a register, a minimum error determination device, and a primary / secondary codebook management device are installed inside the codebook control device 600. After the entry is read out from the head address of the above, the error between the synthesized voice and the voice input from the outside is determined, and the evaluation result is input to the register and stored. The above operation is repeated until the end of the primary codebook 101a is reached. Among the evaluation results stored in the register, a code corresponding to a sound source parameter determined to have the smallest error by the minimum error determination device is output to the outside.

During this time, the internal states of the primary codebook 101a, the secondary codebook 101b, and the codebook control device 600 change sequentially during operation. Next, when the end of the voice encoding process is instructed (step 15c), the primary codebook 1 is stored at that time.
01a, secondary codebook 101b, and codebook control device 60
The internal state of 0 is stored in the personal codebook 601 (step 16c). In this way, the primary codebook 101a optimized during the previous speech encoding process can be used from the beginning. Further, the personal codebook 601
Holds a plurality of states distinguished by a personal identification signal, so that there is no problem even when the speech coding apparatus of the present invention is used by a plurality of persons. Furthermore, if the personal codebook 601 is configured by a nonvolatile storage device such as a ferroelectric memory, the contents of the personal codebook can be retained even when the power is turned off, and high-speed operation can be performed. In addition, when the data is exchanged between the upper layer codebook 101a and the lower layer codebook 101b, the error determination device 400b determines the upper layer codebook based on the history of the voice encoding processing results up to that time. Hierarchical codebook 101a
Are replaced with the data on the codebook 101b of the lower layer. Further, at the end of the speech encoding process, the codebook management device 400c saves the contents of the codebooks 101a and 101b of the upper hierarchy and the lower hierarchy to another storage device (not shown), and stores the next speech encoding. At the start of the processing, the contents of the storage device may be controlled to be read out to the codebooks 101a and 101b of the upper layer and the lower layer.

FIG. 8 is a block diagram of a speech coding apparatus according to a fourth embodiment of the present invention. The feature of the present embodiment is that the address table is held in the codebook control device, and the entries on the codebook 101 to be used are limited to those registered in the address table. It is intended to use electric power. The codebook control device 800 includes:
A register 800a for holding an error evaluation result output from the error evaluation device 301, and a minimum error determination device 800 for determining a minimum value from the values in the register 800a.
b, an address table 800c holding an address of the entry to be read on the codebook 101, and an address table 800c
Is provided with an address table management device 800d for managing the contents of the address table. Then, it controls an entry reading device 302 from the codebook 101 and controls a selecting device 303 for a code output from the codebook 101 to the outside. Address table 8
00c holds a part of the entry storage address of the codebook 101. First, the entry of the codebook 101 corresponding to the address read from the address table 800c is read. Next, speech is synthesized according to a predetermined procedure based on the sound source parameters in the entry.
An error between the synthesized voice and the voice input from outside is determined, and the evaluation result is input to the register 800a and stored. Further, the entry is read from the next address read from the address table 800c, and the same operation as the first is performed. The above operation is repeated until the end of the address table 800c is reached.

By doing so, the address table 800
After the processing of the entry read from the address registered at the end of c is completed, the register 800a stores the evaluation results of all the sound source parameters registered in the address table 800c. From the series of evaluation results, the code corresponding to the sound source parameter determined to have the smallest error by the minimum error determination device 800b is selected as a coded version of the input speech and output to the outside. Note that the minimum error determination device 800
The determination result in b is also sent to the address table management device 800d. The address table management device 800d processes the determination result according to a predetermined procedure, and updates the address registered in the address table 800c based on the result. As described above, in the present embodiment, the entries in the codebook 101 used during the audio encoding process are limited to only those whose addresses are registered in the small-capacity address table. Power consumption can be realized. Furthermore, since the registered contents of the address table are updated during operation and optimized so that the input voice can be coded with a smaller error, the voice quality is managed by managing the voice coding with a small-capacity address table. Does not deteriorate. Further, since the address table only holds a part of the entry addresses of the codebook, it is possible to completely match the contents of the conventional codebook with the contents of the codebook. Absent. Furthermore, since only a small-capacity address table is used without any modification to the codebook, the amount of circuits does not increase.

[0019]

As described above, according to the present invention,
When a speech signal is encoded using a codebook that holds a plurality of pairs of sound source parameters and corresponding codes for synthesizing speech, since the encoding process is terminated without accessing the entire codebook, As a result, it is possible to reduce the calculation amount of the audio encoding process, thereby speeding up the encoding process and reducing the power consumption during the encoding process.

[Brief description of the drawings]

FIG. 1 is a basic configuration diagram of a speech encoding processing device according to the present invention.

FIG. 2 is an operation flowchart of a speech encoding method according to the first embodiment of the present invention.

FIG. 3 is a configuration diagram of a speech encoding processing device according to a first embodiment of the present invention.

FIG. 4 is a configuration diagram of a speech encoding processing device according to a second embodiment of the present invention.

5 is an operation flowchart of the speech encoding processing device shown in FIG.

FIG. 6 is a configuration diagram of a speech encoding processing device according to a third embodiment of the present invention.

FIG. 7 is an operation flowchart of the speech encoding processing device shown in FIG. 6;

FIG. 8 is a configuration diagram of a speech encoding processing device according to a fourth embodiment of the present invention.

[Explanation of symbols]

100: voice encoding processing device; 101: codebook, 102
... codebook control device, 101a ... primary codebook, 101b ...
Secondary codebook, 102, 400, 600, 800 ... codebook control device, 102a ... error determination device, 400a, 800a
... Register, 400b, 800b ... Minimum error determination device
400c: primary / secondary codebook management device, 800c: address table, 800d: address table management device, 300: synthesis filter, 301: error evaluation device, 302: entry reading device, 303: code selection device, 601: personal code Book.

Continuing on the front page (72) Inventor, Koji Yamada 5-2-1, Josuihonmachi, Kodaira-shi, Tokyo Inside Semiconductor Division, Hitachi, Ltd. (72) Miki Takeuchi 5-chome, Josuihoncho, Kodaira-shi, Tokyo No. 1 Hitachi Semiconductor Co., Ltd. Semiconductor Division (72) Inventor Hiroyuki Tanikawa 5--20-1, Kamisumihonmachi, Kodaira-shi, Tokyo Incorporated Hitachi Semiconductor Co., Ltd. (72) Inventor Hidetoshi Sekine Kodaira, Tokyo 5-20-1, Kamizuhoncho Inside Semiconductor Division, Hitachi, Ltd.

Claims (10)

[Claims] 1. A sound source parameter which is sequentially read out from a codebook in which a plurality of pairs of a sound source parameter for synthesizing a voice and a code corresponding to the sound source parameter are registered, and a voice reproduced by synthesizing the sound source parameter and inputted. In a speech coding method for selecting and determining a code by evaluating an error with a speech, when the error evaluation value between the speech read and reproduced from the codebook and the input speech is smaller than a threshold, the code is read out. Or outputs the code corresponding to the sound source parameter, or divides the codebook into two, and sequentially reads out and reproduces the error evaluation value between the input voice and the voice read and reproduced sequentially from only one of the divided codebooks. The code corresponding to the sound source parameter of the smallest one is output, or the sound source parameter read from the codebook is registered in an address table stored separately. By outputting the code corresponding to the sound source parameter of the smallest one of the error evaluation values of the reproduced voice and the input voice by limiting the voice code without accessing the entire codebook, A speech coding method characterized by completing a coding process. 2. A codebook in which a plurality of pairs of a sound source parameter for synthesizing speech and a code corresponding to the sound source parameter are registered, and a sound source parameter read out from the codebook is synthesized to generate a speech. And an error evaluator for evaluating an error between the input speech and the input speech. In a speech encoding processor for selecting and determining a code to be output, a magnitude relationship between an evaluation result value of the error evaluator and a constant is determined. An error determination device; and a code selection device for outputting a code corresponding to a sound source parameter corresponding to an evaluation result value determined to be smaller than a constant in the determination by the error determination device to the outside. Audio coding processor. 3. The speech coding apparatus according to claim 2, wherein, in addition to the error determination apparatus, results determined by the error determination apparatus are held for the number of entries in the codebook, or the code is stored. A storage device that holds a number smaller than the number of entries in the book and sequentially updates them to a minimum value, and outputs a code corresponding to the minimum value among a plurality of determination results stored in the storage device to the outside A speech coding processing device characterized by performing the following. 4. A code book in which a plurality of pairs of sound source parameters for synthesizing speech and codes corresponding to the sound source parameters are registered, and a sound source parameter read out from the code book is synthesized, and a generated speech is generated. And an error evaluator that evaluates an error between the input voice and the input voice. In the voice coding processing device that selects and determines a code to be output, two or more upper layers having a smaller capacity than lower layers A codebook having a hierarchy, a storage device for reading out excitation parameters only from the codebook of the upper hierarchy, and holding an evaluation result of an error between a speech generated by synthesizing the excitation parameters and an input speech; Among the evaluation results stored in the device, the maximum value and the minimum value are determined, the code corresponding to the minimum value is output to the outside, and the code and voice corresponding to the maximum value are output. Pa An error determination device for moving a set of parameters from the upper layer to a codebook of a lower layer; and a codebook management device for managing exchange of data between a codebook belonging to the upper layer and a codebook belonging to a lower layer. And a speech coding processing device comprising: 5. The speech coding apparatus according to claim 4, wherein the error determination apparatus is configured to replace data between an upper-layer codebook and a lower-layer codebook. A code that is determined to have the largest error in the upper-layer codebook based on the history of the voice-encoding processing result of Processing equipment. 6. The speech encoding device according to claim 4, wherein the codebook management device stores the contents of the codebooks of the upper layer and the lower layer in another storage device when the speech encoding process ends. A voice encoding processing device that controls the contents of the storage device to be read into the upper-level and lower-level codebooks at the start of the next audio encoding process. 7. The speech encoding processing device according to claim 4, wherein the speech encoding processing is performed as a storage device that saves the contents of a codebook of a higher hierarchy and a lower hierarchy at the end of the speech encoding processing. An audio encoding processing device, which is a personal codebook in which contents are stored for each user who uses the encoding processing device. 8. The speech encoding processing device according to claim 4, wherein said codebook is constituted by a nonvolatile semiconductor memory device. apparatus. 9. A code book in which a plurality of pairs of sound source parameters for synthesizing speech and codes corresponding to the sound source parameters are registered, and a sound source parameter read out from the code book is synthesized, and a generated speech is generated. And a speech encoding processor for selecting and determining a code to be output by using an error evaluator for evaluating an error between the input speech and a speech. A storage device for holding the evaluation result of the error, and a storage device for storing the error evaluation result. An error determination device that determines the maximum value and the minimum value of the evaluation results obtained, an address table storage device that holds a part of the data storage address of the codebook, and a registration in the address table storage device And a management device for managing addresses to be processed. 10. The speech encoding processing device according to claim 9, wherein said address table storage device mainly holds an address corresponding to a code having a small error evaluation result value. Encoding processing device.