JPH08146997A - Device and system for code conversion - Google Patents

Abstract

PURPOSE: To provide the code conversion system which allows a telephone conversation without reconverting a speech into a reproduced speech temporarily as to a telephone conversation between speech encoding systems that are different in quantized value or quantizing method. CONSTITUTION: When there are a 1st speech encoding method and a 2nd speech encoding method which perform encoding processes differing in quantized value or quantizing method as code excitation linear predictive encoding methods, the code converting device converts multiplexed codes of the 1st speech encoding method into multiplexed codes of the 2nd speech encoding method. A code separation part 510 inputs the multiplexed codes encoded by the 1st speech encoding method and separates them into individual codes, and a conversion part 500 converts the individual separated codes into respective codes of the 2nd speech encoding method according to the correspondence relation between the codes of the 1st speech encoding method and the codes of the 2nd speech encoding method. A multiplexing part 570 multiplexes the respective codes of the 2nd speech encoding method which are converted by the conversion part 500.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a code conversion system equipped with voice code conversion means for converting coded voice used in automobile telephone devices, portable telephone devices, communication equipment and the like.

[0002]

2. Description of the Related Art Today's digital cellular communication systems incorporate advanced voice coding techniques to make effective use of the transmission frequency band. As a speech coding method used in this digital cellular communication system, for example, a vector sum excitation linear predictive coding method (hereinafter referred to as VS
ELP (Vector-Sum Excited Linear Predictive Codin
g) There is a method).

FIG. 21 shows the simple configuration and operation of this system.
Will be described with reference to.

In FIG. 21, a cellular communication system using this VSELP method includes a mobile terminal A100 and a mobile terminal B160 which wirelessly communicate with each other, and a radio base station A110 and a radio base station B150 which wirelessly communicate with the mobile terminal. , A switch A120 and a switch B140 for exchanging calls, and a transmission line 130 for connecting the switches.
And a telephone 170 connected to the exchange.

The mobile terminal A100 and the radio base station A11
0 is an encoding device A10 using the VSELP method.
1. The mobile terminal B16 is equipped with the encoding device A111.
0 and the radio base station B150 have the same method as the coding device B.
161 and an encoding device B 151 are mounted respectively.

The operation of the mobile terminal A100 and the telephone 170 during a call will be described. The audio signal input to the mobile terminal A100 is coded by the coding device A in the terminal, and 1
A transmission code of 1.2 kbps is generated, and the wireless base station A1
It is output to 10. In the wireless base station A110, the received transmission code of 11.2 kbps is decoded from the code in the encoding device A111, and the voice signal is reproduced. The reproduced voice signal is transmitted to the telephone set 170 via the exchange A120, the transmission path 130 and the exchange B140.
In this case, assuming that the encoding / decoding by the VSELP method is one link, the mobile terminal A100 and the mobile terminal B1 are
The call with 60 has a two-link structure, and causes a deterioration in voice quality due to an increase in quantization error and a signal processing delay during encoding.

[0007] As a countermeasure against this, there is a technique of "network control for improving voice communication quality between mobile units" which is a technique for preventing the deterioration of voice quality by connecting the coding device of a radio base station through to establish a digital 1 link between terminals. (Spring 1991 IEICE B-349) is proposed. According to this conventional technique, VSE is transmitted in the network without performing decoding in the radio base station.
By transferring the LP code as it is, the communication quality is improved.

Next, an outline of the VSELP method is described in "VECT
OR SUM EXCITED LINEAR PREDICTION (VSELP) SPEECH COD
ING FOR JAPAN DIGITAL CELLULAR "(The Institute of Electronics, Information and Communication Engineers Radio Communication Systems Research Group, IEICE Technical Report RCS90-26).

This encoding method is composed of two basic processes. One is a voice coding process, and the other is a transmission line coding process. In this encoding method, first, an input voice signal is subjected to a voice encoding process to generate a voice code of 6.7 kbps, and further, a convolutional code is used as a transmission line encoding process in order to provide error resistance in a radio wave transmission section. Apply 1
It generates and outputs a transmission code of 1.2 kbps.

Further, the decoder receives 11.2 kbp.
The transmission code of s is subjected to Viterbi decoding which is a transmission path decoding process in order to correct the bit error generated in the radio wave transmission section,
A voice code of 6.7 kbps is generated, and a voice signal is generated by a voice decoding process based on this code.

The voice encoding process of this VSELP method is 2
It has two excitation sources, one of which is an adaptive codebook and the other is a fixed codebook composed of linear sums of vectors, which is a feature of the VSELP method. The operation of the speech coding process compares an input speech waveform with a synthesized speech waveform generated by passing an excitation signal, which is an addition signal of two excitation sources, through a synthesis filter, and an adaptive codebook that minimizes the error between them. And a code of a fixed codebook and a gain codebook for adjusting the gains of the two are selected. Then, it becomes a mechanism to transmit these codes (codes of adaptive codebook and fixed codebook and gain codebook) together with a code obtained by quantizing the filter coefficient of the synthesis filter and a code obtained by quantizing the power value of the audio signal. There is. The filter coefficient and power value of the synthesis filter are updated every 20 ms, and the codes of the adaptive codebook, fixed codebook, and gain codebook are updated every 5 ms. The voice decoding process is a mechanism that reproduces voice based on each code by using the voice synthesis process of the encoding process.

Further, a pitch-synchronized code excitation linear predictive coding method (hereinafter referred to as PSI-CELP method) which realizes a transmission rate of 5.6 kbps which is half of the VSELP method in order to effectively use the transmission frequency band, is 「Pitch Sync
hronous Innovation CELP (PSI-CELP) ”(IEICE RCS93-78). In this PSI-CELP method, a noise canceller and a low sound volume suppression process are performed on an input voice prior to the voice coding process and the transmission line coding process to improve characteristics in an actual use environment.

The excitation source of the voice coding process in the PSI-CELP method is composed of two parts, one part for switching between the adaptive codebook and the fixed codebook, and the other part for the noise codebook having a two-channel structure. It consists of The main feature is that the output of each noise codebook is synchronized with the cycle corresponding to the output of the adaptive codebook, is cycled, and the codes are added. Further, each excitation source multiplies the gain and then drives the synthesis filter to generate synthetic speech. In the voice encoding process, the synthesized voice is compared with the input voice that has passed through the noise canceller and the low volume suppression, and the code of each codebook having the smallest error is selected. Also, their sign,
This is a mechanism for transmitting a code in which the filter coefficient of the synthesis filter is quantized and a code in which the power value is quantized. The filter coefficient and power value of the synthesis filter are updated every 40 ms, and the codes of the adaptive codebook, fixed codebook, and gain codebook are updated every 10 ms. The 3.45 kbps voice code generated by the voice coding process is subjected to convolutional coding as a transmission line coding process to further provide error resistance in the radio wave transmission section, and then 5.6 kbps.
The transmission code of is generated and output. In the decoder, the received 5.6 kbps transmission code is subjected to Viterbi decoding, which is a channel decoding process, in order to correct the bit error generated in the radio wave transmission section, and a 3.45 kbps voice code is generated. The voice decoding process is a mechanism for reproducing voice based on each code by using the voice synthesis process of the encoding process.

[0014]

In future digital cellular communication systems, the VSELP method and P
Since the speech coding apparatus with the SI-CELP method is standardized, it is expected that both speech coding apparatuses will coexist. As mentioned above, VSELP method and PSI-CE
The LP method is a code-excited linear predictive coding method with a basic configuration, but the code transmission rate is different, and the quantization value or the quantization method is different in each vector quantization. Therefore, a code encoded by the VSELP method is used as a PSI-CEL.
It cannot be decrypted by the P method. Therefore VSEL
In a call between a mobile terminal equipped with the P method and a mobile terminal equipped with the PSI-CELP method, a code coded by one method is once decoded and returned to reproduced voice, and then coded and decoded again by the other method. There is no choice but to convert. This allows
The two-link configuration as described above results in deterioration of voice quality due to increase in quantization error and signal processing delay during encoding.

As described above, according to the conventional technique, VSELP is used.
Digital 1 link is not possible in a call between a mobile terminal equipped with the method and a mobile terminal of the PSI-CELP method. For this reason, it is possible to talk by performing digital 2 link between voice coding methods having different quantized values or quantization methods, but voice quality due to quantization error is generated because voice signals are decoded once and voice coded again. And the signal processing delay is increased.

An object of the present invention is a code-excited linear predictive coding method, which causes deterioration of communication quality in a call between first and second speech coding methods having different quantized values or different quantized methods. It is to provide a code conversion device and a code conversion system which prevent the conversion.

Further, in the code-excited linear predictive coding method, in a call between the first and second voice coding methods having different quantized values or quantization methods, the call is made without returning to the reproduced voice once. It is another object to provide a code conversion device and a code conversion system that can

[0018]

In order to achieve the above object, a speech signal is analyzed by code excitation linear prediction at predetermined analysis time intervals and separated into spectral envelope information and spectral fine structure. A short-time analysis code obtained by quantizing and encoding spectral envelope information, a power code obtained by quantizing and encoding the power of a voice signal, and the spectrum fine structure, and a predetermined prediction lag range of a long-term prediction lag filter. Accordingly, the long-term prediction lag code obtained by quantizing and encoding the filter coefficient of the long-term prediction lag filter and the gain code obtained by quantizing and encoding the amplitude of the spectrum fine structure A multiplex code encoded by a first speech encoding method for encoding is a code excitation linear predictive encoding method, Is a code conversion device for converting into a multiplex code according to a second voice encoding method for performing encoding with a different quantized value from the voice encoding method according to the first voice encoding method. A code separation unit that inputs a coded code and separates each code, and each code separated by the code separation unit is coded by the first speech coding method and the second speech coding. A conversion unit for converting each code according to the second speech coding method into each code according to the correspondence with the code according to the method, and each code according to the second speech coding method converted by the conversion unit. And a multiplexing unit for converting the data.

Further, a first coder for coding by the first speech coding method and a second speech coding method for coding different analysis method from the first speech coding method. A second decoder for decoding and a multiplex code obtained by the first encoder are converted into a multiplex code obtained by the second speech encoding method, and then the second decoder is provided. In a transcoding system having a transcoding device for outputting, whether the coding method of the connection destination of the transcoding device is the first speech coding method or the second speech coding method. Connection destination information storage means for storing the information in advance and the connection destination information storage means, and when the connection destination encoding method is the second speech encoding method,
The code conversion apparatus may further include control means for controlling the code conversion.

Further, in the communication system having the first communication device having such a code conversion device and the second communication device for performing the encoding by the second speech encoding method, the first communication device is provided. Is an inquiring means for inquiring an encoding method of a communication device which is a communication destination, and a response from the communication device of the communication destination to the inquiry by the inquiring means, and the encoding method of the communication device of the communication destination is received by the response. In the case of the second encoding method, the code conversion device converts the multiplex code encoded by the first encoding method into a multiplex code by the second speech encoding method. And a first control means for notifying the communication device of the communication destination that the encoding is not performed by the second encoding method, and the second communication device. A response unit that responds when the inquiry unit inquires about the encoding method by saying that the encoding method in the second communication device is the second encoding method; When the notification that the encoding is not performed by the second encoding method is received, the second communication method is performed by the first communication device without performing the encoding by the second encoding method. Second control means for controlling the signal of the multiplex code by the voice coding method of 1) to be sent to the terminal.

Further, the communication device for encoding the voice signal by the first voice encoding method may include the above-mentioned code conversion device.

[0022]

In the code conversion device, the code separation unit inputs the multiplexed code encoded by the first speech encoding method, separates each code, and the conversion unit uses the code separation unit. Each separated code is converted into each code by the second speech coding method according to the correspondence relationship between the code by the first speech coding method and the code by the second speech coding method. To do. The multiplexing unit multiplexes and outputs each code converted by the conversion unit by the second audio encoding method to output a second code output encoded by the first audio encoding method. Can be directly converted into the code output of the voice encoding method of No. 1, and the code output of the first voice encoding method can be reproduced by the decoder of the second voice encoding method without returning to the reproduced voice once. You can For this reason, it is possible to suppress coding distortion that occurs during digital two-link and provide good communication quality.

For example, the conversion unit is provided with a table showing the correspondence relationship between the code according to the first speech coding method and the code according to the second speech coding method, and by referring to the table, The conversion may be performed.

For example, of the first and second speech coding methods, one is the above-described vector sum excitation linear predictive encoding method (VSELP), and the other is the pitch-synchronous code excitation linear predictive encoding. Method (PSI-CELP). If there are VSELP standardized in Japan and VSELP standardized in the United States as VSELP, these can be used as the first and second voice encoding methods.

Further, the conversion section converts the short time analysis code encoded by the first speech encoding method into the first speech analysis code.
Short-term prediction that is converted into the short-time analytic code by the second speech coding method according to the correspondence relationship between the short-time analytic code by the second speech coding method and the short-time analytic code by the second speech coding method. Analysis code conversion means, and the first
The power code encoded by the voice encoding method according to the above method, according to the correspondence relationship between the power code according to the first voice encoding method and the power code according to the second voice encoding method,
A power code conversion means for converting into a power code by the second speech coding method, and a long-term prediction lag code coded by the first speech coding method, a long-term prediction by the first speech coding method. According to the correspondence between the lag code and the long-term predicted lag code by the second speech coding method,
Long-term prediction lag code conversion means for converting into a long-term prediction lag code by the second speech coding method, and long-term prediction lag code, gain code and noise excitation code conversion code coded by the first speech coding method. According to the first speech coding method, the excitation signal is generated, and the second speech coding is performed according to the long-term prediction lag code by the second speech coding method converted by the long-term prediction lag code conversion means. Generating the excitation signal in the method and minimizing the error between the excitation signal in the first speech encoding method and the excitation signal in the second speech encoding method;
With the gain code and the noise excitation code conversion means for outputting the gain code and the noise excitation code conversion code according to the speech encoding method, the conversion means can perform conversion for each code.

[0026]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows a block diagram of a first embodiment of the code conversion system. In FIG. 1, the code conversion system includes a coder A200 for coding a voice according to a specific voice coding system A, and a coded code according to a specific voice coding system A for a specific voice coding system A. A code conversion device 220 for converting into a code of a voice encoding method B different from
A transmission path A210 for connecting the encoder A200 and the code converter 220, a decoder B240 for decoding a code coded by the voice coding method B, and a decoder B24.
Transmission line 2 for connecting 0 and the code converter 220
30 and 30.

The encoder A200 and the decoder B240 are basically speech coding systems having a code excitation linear predictive coding system as a basic structure, and quantizer values used when quantizing each parameter or It uses a voice coding method with different quantization methods. In this example,
The different speech coding schemes are code excitation predictive coding schemes that both have the same basic configuration, and use speech coding schemes with different quantization values or quantization methods used when quantizing each parameter. And For example, there are a VSELP system used in the full rate voice encoding system of domestic digital cellular mobile phones and a PSI-CELP system used in the half rate voice encoding system. In the present embodiment, these will be described as examples. .

In FIG. 1, the digital voice signal input from Sin is input to the encoder A200 and voice encoded. This code is input to the code conversion device 220 via the transmission path A210. In this code conversion device 220, the received code of the encoder A200 is directly converted into a code that can be used by a decoder B240 that uses a voice encoding method B different from the voice encoding method A used by the encoder A200. Converting. The code converted by the code conversion device 220 is input to the decoder B240 via the transmission path B230. Based on this code, the decoder B240 decodes the digital audio signal. In this way, it is possible to suppress the deterioration of the voice quality by directly converting the code without returning to the voice signal once, and to reduce the scale of the hardware as compared with the case of using two links.

Next, a specific configuration of the code conversion device 220 shown in FIG. 1 will be described. FIG. 2 shows a functional block diagram of the code conversion apparatus according to the present embodiment. In FIG. 2, two transcoding devices are used in a transmission line of an actual communication system to realize bidirectional communication. In order to simplify the description, a one-way code conversion device will be described here.

In FIG. 2, the code conversion device 220 converts a code coded by a specific audio coding method A into a code of a audio coding method B different from the specific audio coding method A. The conversion unit 500, a transmission line decoding unit 580 that performs Viterbi decoding to correct a bit error, and a transmission line encoding unit 590 that performs convolutional coding to provide error resistance of a radio wave transmission section are included. Here, the transmission line coding unit 590 and the transmission line decoding unit 580 are necessary when the code conversion device is used in the wireless communication system of the digital mobile phone. When used, this transmission line coding unit 59
The portion of 0 and the transmission path decoding unit 580 can be unused or not provided.

In the speech code conversion unit 500, noting that both the VSELP system and the PSI-CELP system are code excitation linear predictive coding systems, the codes of the respective parameters (adaptive codebook, fixed codebook, There is a code corresponding to each code of the gain codebook, a code obtained by quantizing the filter coefficient of the synthesizing filter, and a code obtained by quantizing the power value), and a conversion table of the corresponding relationship between the codes is obtained. And memorize each.

In the speech code conversion section 500, a code separation section 510 for separating each code, a gain code fixed codebook code conversion section 520 for performing gain code and fixed codebook code (noise excitation code) conversion, and a short-term prediction analysis parameter code. -Term prediction analysis parameter code conversion unit 5 for converting
40, a power code conversion unit 550 that converts a code obtained by quantizing a power value, a long-term prediction lag code conversion unit 560 that converts a long-term prediction lag code, and a code conversion control unit 5 that controls each conversion unit
30 and a code multiplexing unit 57 for multiplexing each code
Has zero.

The code conversion device will be described below.

In FIG. 2, the code separation unit 510 has a Ci
From n, the multiplexed code that is channel-decoded through the channel decoding unit 580 and is input to the voice code conversion unit 500 is
The code that can be input to each conversion unit is separated into a gain code fixed codebook code, a short-term prediction analysis parameter code, a power code, and a long-term prediction lag code. The code separation unit 510 includes a gain code fixed codebook code conversion unit 520 for the gain code fixed codebook code and a short-term prediction analysis parameter code converter 54 for the short-term prediction analysis parameter code.
0, the power code is sent to the power code conversion unit 550, and the long-term prediction lag code is sent to the long-term prediction lag code conversion unit 560.

Short-term prediction analysis parameter code conversion unit 540
Converts the received short-term prediction analysis parameter code according to the short-term prediction analysis parameter code conversion rule. The power code conversion unit 550 converts the received power code according to the power code conversion rule. The long-term prediction lag code conversion unit 560 converts the received long-term prediction lag code according to the long-term prediction lag code conversion rule. The code conversion control unit 530 receives the long-term predicted lag code before and after the conversion, converts the long-term predicted lag code into a quantized value, and converts the long-term predicted lag code into a gain code fixed codebook code conversion unit 52.
Output to 0. The gain code fixed codebook code conversion unit 520 converts the received gain code fixed codebook code based on the long-term predicted lag values before and after conversion received from the code conversion control unit 530. The code converted by each conversion unit is aggregated in the code multiplexing unit 570 and multiplexed again, and the code passes through the transmission line encoding unit 590 and is subjected to transmission line encoding and output to Cout. As a result, it is possible to directly convert the code without converting the code having a different voice encoding algorithm into the digital voice signal.

For example, a VSELP system voice code and P
When converting a voice code with the SI-CELP system, VS is used.
The vector quantization code of the reflection coefficient parameter of the short-term prediction analysis of the ELP method and the vector quantization code of the line spectrum pair parameter obtained by converting the short-term prediction coefficient parameter of the PLP-CELP method are converted, and the power code of the VSELP method and the PSI- Converts with the CELP system power code to obtain VSE
Long-term predictive lag code and PS of adaptive codebook of LP system
The long-term prediction lag code of the adaptive codebook of the I-CELP method is converted according to a predetermined conversion rule. Also,
Since the PSI-CELP system is further provided with a noise codebook code, an excitation signal is generated from the fixed codebook code and the gain codebook code of the VSELP system, and the fixed codebook and the noise code of the PSI-CELP system are also provided. The excitation signal is generated based on the book and the gain codebook, and the noise codebook code and the gain codebook code in the PSI-CELP method are determined so as to minimize the error of these excitation signals.

Next, details of the conversion rule will be described with reference to the block diagram of each conversion unit.

FIG. 3 shows a functional block diagram of the short-term prediction analysis parameter code conversion unit 540 according to this embodiment. FIG.
In the short-term prediction analysis parameter code conversion unit 540
Is a short-term prediction analysis parameter interpolator 6 for interpolating the short-term prediction analysis parameter code in order to adjust the code transmission rate.
10, a short-term prediction analysis parameter conversion control unit 620 that controls the short-term prediction analysis parameter conversion, and a short-term prediction analysis parameter code corresponding unit 630 that stores a conversion table and performs code conversion with reference to the conversion table.

Short-term prediction analysis parameter code correspondence unit 630
The conversion table of is to compare the quantized values indicated by the short-term prediction analysis parameter codes on both the input side and the output side to be converted,
It can be created by associating both codes so that the error is minimized. In addition, when the method of generating the analysis parameter indicated by the short-term prediction analysis parameter code is different between the two, one analysis parameter is converted into an analysis parameter suitable for the other parameter generation method, and this analysis parameter is used to convert the analysis parameter -Create a conversion rule by associating the codes so that the Crid distance is minimized. When the analysis orders of both short-term prediction analysis parameters are different, the analysis orders are aligned by converting the parameters indicated by both analysis parameter codes into cepstral parameters that can be converted into different orders, and the cepstrum parameters before and after conversion A conversion rule is created by associating the codes so that the distance is minimized.

For example, VSELP system and PSI-CEL
A method of creating a conversion rule of the short-term prediction analysis parameter code with the P method will be described below.

The short-term predictive analysis parameter in the VSELP system speech coding is a 10th-order reflection coefficient or PAR.
PS is obtained by performing quantization with the COR coefficient, and
In the I-CELP method speech coding, quantization is performed by using a 10th-order LSP parameter (line spectrum pair parameter obtained by converting a short-term prediction analysis parameter).
Therefore, in order to make the analysis parameters common, VSELP
The scheme's reflection coefficient parameters are converted to short-term predictive analysis parameters and then to line spectrum pair parameters.
A code conversion rule is created by comparing the analysis parameters of both the short-term prediction analysis parameter and the line spectrum pair parameter, and making the codes correspond to each other so that the Euclidean distance is minimized. In this way, the corresponding code is obtained in advance by calculation, a conversion table of the corresponding code is created, and stored in the short-term prediction analysis parameter code corresponding unit 630 shown in FIG.

Next, the operation of the short-term prediction analysis parameter code conversion section 540 shown in FIG. 3 will be described.

The short-term prediction analysis parameter code separated by the code separation unit 510 shown in FIG. 2 is converted into Di shown in FIG.
It is input from n and received by the short-term prediction analysis parameter code conversion control unit 620. Further, this code is transmitted to the short-term prediction analysis parameter code corresponding section 630. The short-term prediction analysis parameter code corresponding unit 630 performs code conversion on the received code according to the conversion table in the corresponding unit, and transmits the converted code to the short-term prediction analysis parameter code conversion control unit 620. The short-term prediction analysis parameter code conversion control unit 620 inversely converts the converted short-term prediction analysis parameter code into a short-term prediction analysis parameter, and transmits the converted code and the converted parameter to the short-term prediction analysis parameter interpolation unit 610. To do. In the short-term prediction analysis parameter interpolation unit 610,
When the code transmission rates of Din and Dout are different, a code is generated and added or deleted to match the converted rate, and the code transmission rate is adjusted.

Next, the interpolation method of this code will be described.
For example, when converting a short-term prediction partial parameter code obtained by PSI-CELP speech coding into a short-term prediction parameter code of VSELP speech coding, PSI-C
In ELP system voice encoding, a code is transmitted every 40 ms, whereas in VSELP system voice encoding, a code is transmitted every 20 ms. Therefore, a code that can be transmitted every 20 ms is generated from a code that is transmitted every 40 ms, and the code is added. Short-term prediction analysis parameter code conversion control unit 620
However, the codes received from Din as the short-term prediction analysis parameter code and converted by the short-term prediction analysis parameter code corresponding unit 630 are Crn-1 and Crn (n is a natural number, Crn-
1 denotes a code immediately before Crn (40 ms before), and a short-term prediction analysis parameter indicated by the code is α (m is a natural number, α _m is a vector element of m-th order, and α ₀ ^{n to} α _m ⁿ represent vector elements of Crn). It is assumed that the vector of this code and parameter is given by the following equation.

[0046]

[Equation 1]

[0047]

[Equation 2]

Based on this code, the code generated between Cr _n-1 and Crn is set to Cr _n-0.5 (Cr _n-0.5 indicates a code 20 ms before Crn), and this code Cr _n-0.5 is generated. To do. The generation of the code, using the vectors of parameters of the Cr _n-1 and Cr _n, is obtained by the following equation for obtaining an average of parameters between Cr _n-1 and Cr _n.

[0049]

(Equation 3)

As described above, the code Cr _{n -0.5} can be generated and the respective vectors of Cr _{n -1} , Cr _{n -0.5,} and Cr _n shown in Expressions 4 to 6 can be obtained.

[0051]

[Equation 4]

[0052]

(Equation 5)

[0053]

(Equation 6)

On the contrary, when converting the code of the VSELP system voice coding to the code of the PSI-CELP system voice coding, the code CP _{n -0.5} is deleted and transmitted. The added code and the converted code are transmitted to the short-term prediction analysis parameter code conversion control unit 820 and further output to Dout.

As described above, when the transmission rates of the codes are different, the codes can be supplemented by adding or deleting the codes so as to match the transmission rate after conversion.

Next, the power code converter 550 will be described. FIG. 4 is a functional block diagram of the power code conversion unit 550 according to this embodiment. In FIG. 4, the power code conversion unit 750 includes a power value interpolation unit 710 that interpolates power values in order to adjust the code transmission rate, a power code conversion control unit 720 that controls conversion of power codes, and a conversion table. It has a power code corresponding unit 730 that stores and performs code conversion with reference to this conversion table.

The conversion rule of the power code corresponding unit 730 compares the quantized values indicated by the power value codes on both the input side and the output side to be converted, and applies both codes so as to minimize the error. To create. In this case, since the original data used to calculate the power value is different as shown in FIG. 5, the following conversion is performed. In FIG. 5, V
The power values in the SELP method are Rg0, Rg1, and Rg.
It is the scalar value of Rg3 that is quantized and coded with respect to the average signal power of the input signal in the subframe indicated by 2 and Rg3, respectively. Rg0, Rg1,
Since the value of Rg2 is not coded, convert this Rg3 to the average signal power of the input signal in the subframe,
It can be obtained from this average signal power. In addition, in the PSI-CELP system, R0, R1, R
It is the vector quantization of R0, R1, R2, R3 that is quantized and coded for the average signal power of the input signal in the sub-frames denoted by 2 and R3 respectively. As shown in FIG. 5, 1 subframe and 2 subframes in the VSELP system correspond to 3 subframes in the PSI-CELP system. Therefore, when converting from the VSELP method to the PSI-CELP method, the conversion can be performed as shown below.

R2 = (Rg0 + Rg1) / 2 R3 = (Rg2 + Rg3) / 2 Further, R0 and R1 can use N-1 frames of the VSELP system. Also, the reverse can be converted as shown below.

Rg0 = Rg1 = R2 / 2 Rg2 = Rg3 = R3 / 2 In this way, the power code corresponding unit 730 may perform conversion by calculation, or the correspondence obtained by calculation in advance may be converted into a conversion table. You may memorize it.

The power code separated by code separation section 510 shown in FIG. 2 is input from Din shown in FIG. 4 and received by power code conversion control section 920. Further, this code is transmitted to the power code corresponding unit 730. Power code corresponding unit 73
In 0, the received code is code-converted according to the conversion table in the corresponding unit, and the converted code is converted into the power code conversion control unit 720.
Send to. The power code conversion control unit 720 converts the converted power code into a power value, and transmits the converted code and power value to the power value interpolation unit 710. Power value interpolation unit 71
At 0, when the code transmission rates of Din and Dout of the power value interpolation unit 710 are different, a code is generated and added or deleted to match the conversion rate, and the code transmission rate is adjusted.

For example, when converting a power code obtained by PSI-CELP speech coding into a power code for VSELP speech coding, the code is transmitted every 40 ms in PSI-CELP speech coding. In VSELP audio coding, a code is transmitted every 20 ms. Therefore, a code that can be transmitted every 20 ms is generated from a code that is transmitted every 40 ms, and the code is added. According to the above-mentioned power value conversion method, as shown in FIG. 5, the conversion is performed in consideration of the transmission rate of the code.
In the case of 0, the code may be transmitted according to the transmission rate on the output side without newly generating the code.

When the code is converted without considering the transmission rate of the code, the power value interpolating unit 710 may generate the code and adjust the transmission rate as shown below. Good.

The power code conversion control unit 720 receives the codes received from Din as the power code and converted by the power code corresponding unit 730 as CP _n-1 and CP _n (n is a natural number, CP _n-1.
Denotes a code one bit before CP _n (40 ms before), and the power value indicated by the code is represented by a scalar of P (P _n represents a scalar value when CP _n ). It is assumed that the sign and the scalar value are given by the following expressions.

[0064]

(Equation 7)

[0065]

(Equation 8)

Based on this code, the generated code between CP _n-1 and CP _{n is set} to CP _n-0.5 (CP _n-0.5 indicates a code 20 ms before CP _n ) and this code CP _n-0.5 To generate. The generation of this code uses the scalar values of CP _n-1 and CP _n ,
It is obtained by the following formula for averaging the parameters of CP _n-1 and CP _n .

[0067]

[Equation 9]

As described above, it is possible to generate a code and obtain the scalar values of CP _n-1 , CP _n and CP _{n-0.5 shown in} equations 10 to 12.

[0069]

[Equation 10]

[0070]

[Equation 11]

[0071]

(Equation 12)

On the contrary, when converting the code of the VSELP system voice coding to the code of the PSI-CELP system voice coding, the code CP _{n -0.5} is deleted and transmitted. The added code and the converted code are the power code conversion control unit 72.
It is transmitted to 0 and output to Dout.

Next, the long-term prediction lag code conversion section 560 will be described with reference to FIG. FIG. 6 is a functional block diagram of the long-term prediction lag code conversion unit 560 according to this embodiment. In FIG. 6, the long-term prediction lag code conversion unit 56
0 is a long-term prediction lag interpolation unit 810 that interpolates a long-term prediction lag to adjust the code transmission rate, a long-term prediction lag code conversion control unit 820 that controls conversion of the long-term prediction lag code,
Further, it has a long-term prediction lag code corresponding unit 830 which stores a conversion table and performs code conversion with reference to this conversion table.

First, the long-term predicted lag code corresponding section 830 will be described. The range of long-term predicted lag code is VSELP
Since the system and the PSI-CELP system have different cases and same cases, the long-term prediction lag code conversion is performed in consideration of each case. As shown in FIG.
The long-term filter controlled by the long-term prediction lag is characterized in that n of Z ⁻ⁿ (long-term prediction lag) is variable. In this embodiment, the range of the long-term prediction lag code of the voice encoder A200 shown in FIG. 1 which is the VSELP system is Lami.
From n to Lamax, the long-term prediction lag code is La, and the range of the long-term prediction lag of the PSI-CELP speech decoder B240 is from Lbmin to Lbmax,
Let the long-term predicted lag code be Lb. Also, Lamin
And Lbmin, or the relationship between Lamax and Lbmax has the relationship shown by the following equation. here,
N is an integer of 2 or more, and is a value that is arbitrarily determined.

[0075]

(Equation 13)

[0076]

[Equation 14]

[0077]

(Equation 15)

[0078]

[Equation 16]

Under the above conditions, the long-term prediction lag code La of the VSELP type speech coder A200 is changed to PSI-CELP.
Long-term prediction lag, code Lb _{n of the} B-system speech decoder B240
In the case of converting to, the long-term prediction lag code corresponding unit 830 is created based on the correspondence rule described below. Alternatively, an equivalent calculation is performed to perform code conversion.

Next, the algorithm of the correspondence rule in this embodiment will be described with reference to FIG. FIG. 8 shows an algorithm of the correspondence rule. Here, Lb _n-1 indicates the long-term prediction lag code one sample before Lb _n .

First, La is compared with Lbmin (step 910), and if La <Lbmin, step 950.
Then, the value of Lb _n is determined to be N times La (step 950). Here, N is a fixed value of N shown in Expression 13, and an optimum value is obtained by an experiment. If La <Lbmin is not satisfied in step 910, that is, if La ≧ Lbmin, then La> Lbmax is further compared (step 920). La>
In the case of Lbmax, the value of Lb _n is set to the value of Lb _n-1 immediately before (step 930). If La> Lbmax is not satisfied, the value of Lb _n is determined as the value of La (step 940). Performs processing in each condition, after determining the value of Lb _n, the determined Lb _n, and Lb _n-1 in the process of conversion of the next sample value, previously stored in the delay buffer. Lb _n-1 in step 930 described above can fetch a value from this delay buffer.

For example, Lbmin = 20, Lbmax =
If La = 14 and N = 2 at 146,
In step 910, it is determined that La <Lbmin, and the process proceeds to step 950 to determine the value of Lb _n to be 28, which is twice La. N is a value for clearing the condition of the converted minimum value Lbmin. Here, N = 2 is due to the periodicity of the voice pitch of the voice signal as shown in FIG. Therefore, N may be an integer of 2 or more. In FIG. 9, the pitch period of the voice waveform is shown, where N
Since the same waveform is repeated for each point,
It can be seen that there is no problem even if the repetition period is 2N points.

As shown in FIG. 8, the long-term prediction lag code is determined so that it falls within the range of the long-term prediction lag code after conversion, and is converted into a corresponding code according to the encoding of the long-term prediction lag code.

Next, the operation of the long-term prediction lag code conversion section 560 shown in FIG. 6 will be described.

The long-term prediction lag code separated by the code separation unit 510 shown in FIG. 2 is input from Din and received by the long-term prediction lag code conversion control unit 820. Furthermore, this code is transmitted to the long-term prediction lag code corresponding unit 830. The long-term prediction lag code corresponding unit 830 performs code conversion on the received code according to the conversion table in the corresponding unit, and transmits the converted code to the long-term prediction lag code conversion control unit 820. The long-term predicted lag code conversion control unit 820 converts the converted long-term predicted lag code into a long-term predicted lag value, and the converted long-term predicted lag code and the converted long-term predicted lag value in a long-term predicted lag interpolating unit 810. Send. In the long-term prediction lag interpolation unit 810, D
When the code transmission rates of in and Dout are different from each other, the code is generated and added or deleted to match the conversion rate, and the code transmission rate is adjusted.

Next, the interpolation method in the long-term predicted lag interpolation section 810 will be described. For example, when converting a long-term predicted lag code obtained by PSI-CELP speech coding into a long-term predicted lag code for VSELP speech coding, PS
In the I-CELP system voice encoding, the code is transmitted every 40 ms, whereas in the VSELP system voice encoding, the code is transmitted every 20 ms. Therefore, a code that can be transmitted every 20 ms is generated from a code that is transmitted every 40 ms, and the code is added. FIG. 10 shows a long-term predicted lag code in the VSELP system and a long-term predicted lag code in the PSI-CELP system. The long-term prediction lag code in the VSELP scheme is La ₀ , in each subframe,
Represented by La ₁ , La ₂ and La ₃ , respectively. Also,
Also in the PSI-CELP system, Lb ₀ , Lb ₁ , Lb
₂ and Lb ₃ , respectively, but the code transmission rate is different from that of the VSELP system as described above. The long-term prediction lag code conversion control unit 820 receives the codes received from Din as the long-term prediction lag code in the PSI-CELP system and converted into the VSELP system by the long-term prediction lag code corresponding unit 830 as CLa _n , CLa _n-1 (n Is a natural number, CLa
_(n-1 indicates the code immediately before CLa _n (40 ms before))
And the long-term prediction lag indicated by the code is La (La _n is
The vector element is shown). It is assumed that this code and vector are given by the following equation.

[0087]

[Equation 17]

[0088]

(Equation 18)

Based on this code, the generated code between CLa _n-1 and CLa _n is CLa _n-0.5 (CLa _n-0.5 is 2 of CLa _n ).
The code CLa _n-0.5 is generated. The generation of this code is obtained by the following equation.

[0090]

[Formula 19]

[0091]

(Equation 20)

Here, VSELP system and PSI-CE are used.
Since the transmission rate of the long-term prediction lag code is different from that of the LP system as shown in FIG. 10, the code CLa is converted from the code after conversion (the value on the right side in Expressions 19 and 20).
_n-0.5 (left side shown in Expression 19) is generated, and CLa _{n is} further generated.
Is interpolated (left side shown in Expression 20). From the above, the code CLa _n-0.5 and the code CLa _n are expressed by the equations 21 and 2.
2 can be obtained.

[0093]

[Equation 21]

[0094]

[Equation 22]

Further, in the case of converting the code of the VSELP system speech coding into the code of the PSI-CELP system speech coding, the code inputted from Din is first received from Din as the long-term prediction lag code in the VSELP system,
The codes converted into the PSI-CELPP method by the long-term prediction lag code corresponding unit 830 are CLb _n and CLb _n-1 (n is a natural number, CLb _n-1 is one before CLb _n (40 ms before).
, And the long-term prediction lag indicated by the code is L
It is represented by a vector of b (Lb _n represents a vector element).

[0096]

(Equation 23)

[0097]

[Equation 24]

Based on this code, the code is deleted according to the rule shown in the following equation.

[0099]

(Equation 25)

Here, b and c represent vector elements to be compared. In this equation 25, b and c are compared, the larger one is substituted for a, and the value substituted for a is left and other Is to be deleted. Alternatively, as shown in Expression 26, a rule given by the following equation for comparing b and c and substituting the smaller one may be used.

[0101]

(Equation 26)

Using this equation, the elements L _n of the new vector after interpolation are obtained by the equations 23 and 24 as shown in the equations 27 to 30.

[0103]

[Equation 27]

[0104]

[Equation 28]

[0105]

[Equation 29]

[0106]

[Equation 30]

[0107]

[Equation 31]

CLb _n is updated and CLb _n-0.5 is deleted from the generated vector as shown in Expression 31. The converted code is transmitted to the long-term prediction lag code conversion control unit 820 and output to Dout.

As described above, when the transmission rates of the codes are different, it is possible to interpolate the codes so as to match the transmission rate after conversion by adding or deleting the codes.

Next, the gain code fixed codebook code conversion unit 520 will be described with reference to FIG. FIG.
1 shows a functional block diagram of the gain code fixed codebook code conversion unit 520 according to the present embodiment. In the gain code fixed codebook code conversion, an excitation signal is generated by the adaptive codebook and the fixed codebook in each of the VSELP method and the PSI-CELP method, and the error of the excitation signal generated in each method is reduced. The gain code fixed codebook code after conversion is determined so as to be the minimum. Here, in order to reduce the calculation amount, the conversion is performed by generating the excitation signal, but the short-term predictive analysis parameter may be used to synthesize the voice signal to perform the conversion.

In FIG. 11, a gain code fixed codebook code conversion unit 520 includes an input excitation signal generation unit 1000 generated by an input encoding method and an output side excitation signal generation generated by the converted encoding method. Part 101
0, an error calculation unit 1030 that calculates an error between signals generated by the input excitation signal generation unit 1000 and the output side excitation signal generation unit 1010, and an error calculated by the error calculation unit 1030 is minimized. , A minimum error code selection unit 1020 for selecting a fixed gain code fixed codebook code according to the coding method after conversion.

In FIG. 11, the input side excitation signal generator 1
000 is an adaptive codebook A1001 and a fixed codebook A1002 according to the input encoding method, an amplifier A1003 and an amplifier B1004 that amplify a signal.
And a gain codebook A1005 for controlling the amplification factors in the amplifier A1003 and the amplifier B1004.
Have and. Further, the output side excitation signal generation unit 1010 is
Adaptive codebook B1011 according to the coding method after conversion
And fixed codebook B1012, amplifier C1013 and amplifier D1014 for amplifying signals, and amplifier C
1013 and a gain codebook A1005 for controlling the amplification factor in the amplifier D1014.
For example, when converting from the VSELP system to the PSI-CELP system, the input side excitation signal generation unit 1000
An excitation signal is generated by the VSELP method, and the output side excitation signal generation unit 1010 generates the excitation signal by the PSI-CELP method. When the excitation signal is generated by the PSI-CELP method, it further has a noise source codebook, and based on this noise source codebook,
The noise excitation code is determined so that the error between the excitation signals on the input side and the output side is minimized.

The operation of the configuration shown in FIG. 11 will be described. FIG.
1, the long-term prediction lag code separated by the code separation unit 510 shown in FIG. 2 is input to D1in and received by the adaptive codebook A1001. Adaptive codebook A10
At 01, an excitation signal corresponding to the input long-term predicted lag code is output. The fixed codebook code separated by the code separation unit 510 shown in FIG. 2 is input to D2in and received by the fixed codebook A1002, and the fixed codebook A1002 outputs an excitation signal corresponding to the fixed codebook code. To do. The gain code separated by the code separation unit 510 shown in FIG. 2 is input to D3in and received by the gain codebook A1005.
In the gain codebook A1005, the excitation signal from the adaptive codebook A1001 and the fixed codebook A100 are used.
2 is performed using the control of amplification with the excitation signal from 2. Then, the excitation signals of both the amplifier A1003 and the amplifier B1004 are added to generate the excitation signal EXA.

On the other hand, in the output side excitation signal generation section 1011 the long-term prediction lag code converted in the long-term prediction lag code conversion section 560 shown in FIG.
Input to D4in via 30, adaptive codebook B
It is received at 1011. Adaptive codebook B1011 outputs an excitation signal corresponding to the converted long-term prediction lag code. The fixed codebook B1014 outputs the excitation signal according to the value selected by the minimum error code selection unit 1020, and the gain codebook B1015 is
The amplification factors of the amplifier C1013 and the amplifier D1014 are controlled according to the value selected by the minimum error code selection unit 1020. After that, amplifier C1013 and amplifier D10
The excitation signals EXB and 14 are added to generate the excitation signal EXB.

Further, the excitation signal EXA generated by the input side excitation signal generation section and the excitation signal EXB generated by the output side excitation signal generation section 1010 are compared, and the error calculation section 1030 is compared.
Calculate those errors with. Minimum error code selection unit 102
At 0, the fixed codebook B1 is set so that the error is minimized.
014 and the gain codebook B1015 are selected, and the fixed codebook code after conversion to D1out and the gain code to D2out are output. The method of selecting the optimum gain code and fixed codebook code that minimize the error is similar to the case of controlling the input speech and the excitation signal according to the prior art so as to minimize the error, and Instead of comparing with the signal, it can be obtained by comparing the excitation signal EXA and the excitation signal EXB.

By this means, it is possible to output the fixed codebook code and the gain code after conversion.

In the code multiplexing unit 570 shown in FIG. 2, the fixed codebook code and the gain code converted by the gain code fixed codebook code conversion unit 520 and the short-term prediction analysis converted by the short-term prediction analysis parameter code conversion unit 540. The parameter code, the power value converted by the power code conversion unit 550, and the long-term prediction lag code converted by the long-term prediction lag code conversion unit 560 are multiplexed by the multiplexing method of the coding system after conversion and transmitted. The path coding unit 590 performs convolutional coding in order to provide error resistance in the radio wave transmission section. By processing in this way, the code encoded by the specific audio encoding method A can be converted into the code of the audio encoding method B different from the specific audio encoding method A.

FIG. 12 shows an example of a hardware configuration for realizing the above-mentioned code conversion system. In FIG. 12, code conversion dedicated hardware 1500 stores a DSP (digital signal processor) 1510 that performs conversion processing by the voice code conversion unit in the code conversion system described above, a processing program, input data, processing results, and the like in DSP 1510. Memory 15
20 and can have an interface with a host controller if necessary. In FIG. 12, the code input from Cin is subjected to the processing shown in the above-described embodiment by the DSP 1510 to convert the code.
can be output to t. Also, this DSP 1500
May be controlled by the host controller.

By doing so, the hardware scale can be further reduced.

The detailed functions of the code conversion apparatus have been described above.

Next, a digital radio system using the code converter will be described as a second embodiment. In this embodiment, code conversion is performed in the code conversion device, and the code is transmitted without returning to a voice signal by decoding. Therefore, the other side is notified that the link is a digital link, and the other side encodes the code. I try not to do it.

FIG. 13 shows a block diagram of a digital radio system having different voice coding systems. 13, this digital wireless system includes a mobile terminal A300 and a mobile terminal B340 that wirelessly communicate with each other, a wireless base station A310 and a wireless base station B330 that wirelessly communicate with the mobile terminal, and a switch 320 that exchanges a call. Have.

The mobile terminal A300 is an encoding device A301.
Have. The radio base station A310 includes a code conversion device 311 and a coding device A312. The radio base station B330 has an encoding device B331. The mobile terminal B340 has an encoding device B341. Here, the encoding device 301 of the mobile terminal A300 and the encoding device 312 of the wireless base station A310.
Is a voice encoding device using the same voice encoding method,
For example, an encoding device equipped with PSI-CELP system audio encoding or an encoding device equipped with VSELP system audio encoding may be used. Encoding device 33 of radio base station B330
1 and the encoding device 341 of the mobile terminal B340 are speech encoding devices using the same audio encoding method, and the portable terminal A3
00 and a speech coding method different from the speech coding method of the coding apparatus mounted in the radio base station A. For example, if the PSI-CELP system voice coding is installed in the voice coding system of the coding device A301 of the mobile terminal A300 and the VSELP system voice coding is installed in the voice coding system of the coding device B341 of the mobile terminal B340, different voices will be produced. It becomes a communication system between encoding systems. Alternatively, the mobile terminal A300 may use VSELP audio encoding and the mobile terminal B34
0 may be equipped with PSI-CELP audio coding.

FIG. 14 shows the procedure of the procedure before the call start of the system in FIG. In FIG. 14, a procedure for connecting from the mobile terminal A300 to the mobile terminal B340 is shown. The procedure for connecting from the mobile terminal B340 to the mobile terminal A300 can be performed in the same manner.

In FIG. 14, the radio base station A310 is
When there is a call request from the mobile terminal A300, the exchange 32
Through 0, the wireless base station B 330 issues a wireless base station B encoding method request 401 for inquiring about the wireless base station B encoding method (step 401). In the wireless base station B 330, the wireless base station B encoding method request 401
When the mobile terminal B 340 receives the voice coding method, after confirming the voice coding method of the mobile terminal B 340, in order to notify the voice coding method, the radio base station A 310 is added with the identification information of the voice coding method of the mobile terminal B 340. Station B coding system response 40
2 is transmitted (step 402). Radio base station A310
After receiving the wireless base station B encoding method response 402, detects that the voice encoding method of the mobile terminal B 340 is different from the voice encoding method of the mobile terminal A 300, and suppresses decoding by the encoding device A. , A conversion instruction is issued by the code conversion device 311 so as to perform code conversion (step 4).
03). Furthermore, the wireless base station A310 is connected to the mobile terminal B3.
Since the voice coding system of 40 is different from the voice coding system of the mobile terminal A300, a mode for prohibiting the coding by the coding device 331 mounted in the radio base station B330 is set for the radio base station B330. In order to do so, an encoding device B use prohibition request 404 is transmitted (step 40).
4). When the wireless base station B303 receives the use prohibition request for the encoding device B, the wireless base station B303 suppresses the encoding by the encoding device B311 and transmits the code from the exchange 320 to the mobile terminal B340 as it is. Set a certain digital through mode (step 40
5). Furthermore, the wireless base station B303 is connected to the wireless base station A3.
The digital through mode setting, which is a signal indicating that the digital through mode has been set in response to the use prohibition request 404 of the encoding apparatus B, is sent to the terminal 10 (step 40).
6) The usage permission 408 is sent to the mobile terminal B340 (step 408). On the other hand, the radio base station A310, after receiving the digital through mode setting from the radio base station B330, sends the use permission to the mobile terminal A300 (step 407).

Thus, between the radio base stations,
It is possible to confirm the encoding method, and if it is necessary to perform code conversion, the code conversion can be set to be performed, and the decoding can be set not to be performed. By making such a confirmation, it is possible to transmit even between mobile terminals having different audio encoding methods without forming a digital 2 link.

Next, a third embodiment will be described. In the third embodiment, a code conversion device for converting to both VSELP systems in a digital cellular communication system having a Japanese standard VSELP system and an American standard VSELP system as a voice encoding system will be described.

FIG. 15 shows a system provided with a code conversion device and capable of communicating between Japan and the United States by both digital cellular terminals. Further, FIG. 16 shows an example of comparison of frame bit allocation between the Japanese standard VSELP system and the American standard VSELP system. As shown in FIG. 16, the Japanese standard and the American standard have different bit allocations for each code frame, and the American standard has one more codebook. Therefore, since the bit allocation to the frame depends on the transmission bit rate on the transmission path, it can be performed by adjusting the bit rate by the interpolation section of each conversion section in the first embodiment so as to match the converted bit rate. . Also, since there is one more codebook in the American standard,
Gain code fixed codebook code conversion unit 5 shown in FIG.
As in FIG. 20, as shown in FIG. 17, the excitation signal on the input side and the excitation signal on the output side are codebook 17 of the Japanese standard.
00 and American Standard Codebooks 1710 and 1720
And the code of the codebook is selected so that the difference between the two excitation signals is minimized in the minimum error selection unit 1730.

In FIG. 15, this system has a mobile terminal A1100 and a mobile terminal B11 which communicate wirelessly.
90 and a wireless base station A1 that wirelessly communicates with a mobile terminal.
120 and a radio base station B1180, a switch A1130 and a switch B1170 for exchanging calls, a multiplexer A1140 and a multiplexer B1160 for multiplexing or separating data, and a transmission line 1150 for connecting the multiplexers. .

The mobile terminal A1100 is a Japanese standard VSE.
Encoding device A1 for encoding / decoding by the LP system
101, and the wireless base station A1120 is a Japanese standard V
Encoding device A1121 for encoding / decoding by SELP system, and code conversion device 11 for performing code conversion between Japanese standard VSELP system and American standard VSELP system.
22. Further, the mobile terminal B1190 has an encoding device B1101 that performs encoding / decoding according to the American standard VSELP system, and the wireless base station B1180 has an encoding device that performs encoding / decoding according to the American standard VSELP system. It has B1181.

In FIG. 15, the code coded by the coding device A1101 of the portable terminal A1100 by the Japanese standard VSELP system is converted into a code that can be used by the US standard VSELP system by the code conversion device 1122 of the radio base station A1121. After being converted, the exchange A1130, the multiplexer A1140, the transmission line 1150, and the multiplexer B1160.
And the radio base station B1180 via the exchange B1170.
Be transmitted to. The radio base station B1180 directly transmits the code received from the exchange B1170 to the mobile terminal B1190 without using the encoding device B1181. The mobile terminal B1190 reproduces an audio signal from the received code according to the American standard VSELP system.

According to the present embodiment, code conversion can be realized and communication can be performed in a digital cellular communication system provided with the Japanese standard VSELP system and the American standard VSELP system which are different in voice coding. it can.

Next, a fourth embodiment will be described.
FIG. 18 shows a configuration diagram when the code conversion device is applied to a wired communication system.

In FIG. 18, the wired communication system has an exchange A1200 for exchanging calls, an exchange B1210, an exchange C1220 and an exchange D1260, and telephones A1230 and B1240 as communication terminals.

Further, the exchange A1200 is the encoding device A1.
211, and the exchange C1220 has an encoding device B122.
1 and the exchange D1260 has an encoding device A1251. The exchange B1210 includes a code conversion device 1211,
It has connection destination information 1212 that stores the route information and the encoding system of the connection destination. Encoding device A1201 in exchange A1200 and encoding device B12 in exchange D1250
51 is equipped with the audio encoding system A. Switch C
The coding device B 1221 in 1220 is equipped with a speech coding system B different from the speech coding system A. The type of voice coding system installed in each exchange is the connection destination information 12
It is registered in advance in 12.

In FIG. 18, when a telephone call is made between the telephone set A1230 and the telephone set B1240, the voice inputted from the telephone set A1230 is encoded by the encoding apparatus A1 in the exchange A1200.
The voice is encoded in 201, and the code is transmitted to the exchange B1210. The exchange B1210 is a connection destination exchange C12.
For the 20 coding systems, the connection destination information 1212 is referred to, it is detected that the coding systems are different, and it is decided to convert the coding system using the code conversion device 1211. The code from the exchange A1200 is the code conversion device 1 in the exchange B.
At 211, the code is converted into the code of the coding method B that can be used by the coding device B1221 in the exchange C1220.
To 220. The encoding device B1221 in the exchange C1220 decodes the received code into a voice signal and sends it to the telephone set B.

On the other hand, telephone A1230 and telephone C126
When making a call between 0s, the exchange B 1210 detects that the voice encoding method is the same by referring to the connection destination information 1212, and since the encoding method is the same, the exchange B 1210 does not use the code converter 1211. The code from A1200 is sent to the exchange D1250 as it is. Exchange D1
Encoding device A1251 in 250 decodes the received code into a voice signal and sends it to telephone C1260.

According to this embodiment, the code conversion device can be applied to a wired communication system.

Next, a fifth embodiment of the wired communication system using the code conversion device and the multiplexing device will be described. In the present embodiment, the case where the fourth embodiment is further provided with a multiplexer and the multiplexer is provided with a code conversion device will be described as an example.

FIG. 19 shows a block diagram of the wired communication system in this embodiment. In FIG. 19, the wired communication system includes a telephone A 1300 and a telephone B which are communication terminals.
1370 and telephone C1380, and exchange A1310, exchange B1360 and exchange C134 for exchanging calls.
0 and a multiplexer A1 for multiplexing or separating data
320, multiplexer B1330 and multiplexer C13
50 and.

The multiplexer A1320 has an encoder A1321, the multiplexer B1330 has a code converter 1331, and the multiplexer C1350 has an encoder B1351. The encoding device A1321 in the multiplexing device A1320 is equipped with the audio encoding method A.
The coding device B1351 in the multiplexing device C1350 is equipped with a voice coding system B different from the voice coding system A.
In the communication between the multiplexers, since the connection path is fixed, it is prescribed in advance whether to convert the encoding method according to the encoding method of the connection destination or to transmit it as it is. In the example shown in FIG. 19, telephone A130
0 and telephone B1370, the exchange A13
10, multiplexer A 1320, multiplexer B 1330,
It is assumed that the connection path is defined so as to communicate via the multiplexer C1350 and the exchange B1360.

In FIG. 19, when a telephone call is made between the telephone set A1300 and the telephone set B1370, the voice signal input from the telephone set A1300 is transmitted to the multiplexer A1320 through the exchange A1310. The transmitted audio signal is
Speech coding is performed by a coding device A1321 in the multiplexing device A1320. The code is converted in the code converter 1331 of the multiplexer B1330 into a code of the coding method B that can be used by the encoder B1351 in the multiplexer C1350, and the code is sent to the multiplexer C1350. In the multiplexer C1350, the received code is decoded into a voice signal by the encoder B1351 and the exchange B136 is used.
0 to the telephone set B1370.

Further, the telephone set A1300 and the telephone set B137
When making a call between 0s, the exchange A 1310, the multiplexer A
1320, multiplexer B1330, exchange C1340,
A connection path may be defined so that communication is performed via the multiplexer B1330, the multiplexer C1350, and the exchange B1360. In this case, in FIG. 19, when a call is made between the telephone A1300 and the telephone B1370,
The voice signal input from the telephone A1300 is the same as that of the exchange A.
It is transmitted to the multiplexer A 1320 through 1310. The transmitted voice signal is transmitted to the exchange C1340 and input again to the multiplexer B1330. Multiplexer B1
The code conversion device 1331 of 330 converts the input code into a code of the coding method B that can be used by the coding device B1351 in the multiplexing device C1350, and sends the code to the multiplexing device C1350. Multiplexer C1350
Then, the received code is decoded into a voice signal by the encoding device B1351 and transmitted to the telephone set B1370 through the exchange B1360.

Even when such a route is defined, it is possible to use the code conversion device to convert to the encoding system of the output destination.

Next, a sixth embodiment in which a code conversion device is applied as an adapter to a portable terminal will be described. Figure 2
0 is a block diagram of this embodiment.

In FIG. 20, this system has a mobile terminal A1400 that wirelessly communicates, a radio base station A1420 that wirelessly communicates with the mobile terminal, and an exchange 1430 that exchanges a call. The mobile terminal A1400 has an encoding device A1401, the adapter 1410 has a transcoding device 1411, and the radio base station A1420 has an encoding device B1421. The coding device A1401 in the mobile terminal A1400 is equipped with a speech coding system A, and the coding device B1421 in the radio base station A1420 is equipped with a speech coding system B different from the speech coding system A. For example, the encoding device B1421 in the radio base station A1420.
PSI-CELP audio coding is installed in the mobile terminal A1400, and VSELP is applied to the coding device A1401 in the mobile terminal A1400.
When the system voice coding is installed, the code converter 1411 in the adapter 1410 performs code conversion between the PSI-CELP system voice coding code and the VSELP system voice coding code. Whether or not the code conversion is performed may be confirmed by making an inquiry as shown in FIG.

According to this embodiment, this adapter 1410
By using, a mobile terminal equipped with VSELP audio coding can be used even within the PSI-CELP audio coding range.

According to each of the above-described embodiments, both codes are directly converted based on a code conversion rule prepared in advance without returning the code between the audio coding methods having different quantized values or quantizing methods to the audio signal once. A code conversion system for conversion can be constructed. According to this code conversion system, a code conversion unit is added as a hardware scale, but the short-term prediction analysis, power calculation, and voice synthesis can be simplified to prevent an increase in device size and to perform these processes. Since it is reduced, the delay of the signal processing can be improved and the deterioration of the voice quality due to the quantization error can be suppressed.

[0149]

As described above, according to the present invention, it is possible to realize a code conversion system capable of performing a call without returning to the reproduced voice once in a call between voice coding systems having different quantized values or quantization methods. In addition, the short-term prediction analysis, power calculation, and voice synthesis have been simplified, preventing an increase in device scale,
Since these processes are reduced, the delay of signal processing can be improved and the deterioration of the voice quality due to the quantization error can be suppressed.

[Brief description of drawings]

FIG. 1 is a system diagram of a first embodiment of the present invention.

FIG. 2 is a block diagram of a code conversion device according to a first embodiment of the present invention.

FIG. 3 is a functional block diagram of a short-term prediction analysis parameter code conversion unit according to the first example.

FIG. 4 is a functional block diagram of a power code conversion unit according to the first example.

FIG. 5 is an explanatory diagram of power code conversion according to the first embodiment.

FIG. 6 is a functional block diagram of a long-term prediction lag code conversion unit according to the first example.

FIG. 7 is a functional block diagram of an adaptive filter according to the first embodiment.

FIG. 8 is a flowchart of long-term prediction lag code conversion according to the first embodiment.

FIG. 9 is a waveform diagram of an audio signal according to the first embodiment.

FIG. 10 is an explanatory diagram of long-term prediction lag code conversion according to the first embodiment.

FIG. 11 is a functional block diagram of a gain code fixed codebook code conversion unit according to the first embodiment.

FIG. 12 is a hardware block diagram of a code conversion system according to the first embodiment.

FIG. 13 is an explanatory diagram of an application system of the code conversion system according to the second embodiment.

FIG. 14 is a sequence diagram of an operation start of a wireless communication system using the code conversion system according to the first example.

FIG. 15 is an explanatory diagram of an application system of the code conversion system according to the third embodiment.

FIG. 16 is a VSEL between Japan and the US according to the third embodiment.
It is explanatory drawing of the code | symbol of P.

FIG. 17 is a functional block diagram of a gain code fixed codebook code conversion unit according to the third embodiment.

FIG. 18 is an explanatory diagram of an application system of the code conversion system according to the fourth embodiment.

FIG. 19 is an explanatory diagram of an application system of the code conversion system according to the fifth example.

FIG. 20 is an explanatory diagram of an application system of the code conversion system according to the sixth embodiment.

FIG. 21 is an explanatory diagram of a conventional digital cellular communication system.

[Explanation of symbols]

100/300/1100 ... Mobile terminal A, 101/11
1 ・ 301 ・ 312 ・ 1101 ・ 1121 ・ 1201 ・
1321, 1401 ... Encoding device A, 110, 310.
1120 ... Radio base station A, 120/1130/1200
1310 ... Switch A, 130/1150 ... Transmission line, 1
40/1170/1210/1360 ... Exchange B, 15
0 ・ 330 ・ 1180 ... Radio base station B, 160 ・ 340
1190 ... Mobile terminal B, 161.151.331.3.
41/1181/1191/1221/1251/13
51.1421 ... Encoding device B, 200 ... Encoder A, 2
10 ... Transmission line A, 220, 311, 1122, 1211
1331 1411 ... Code conversion system, 230 ... Transmission line B, 240 ... Decoder B, 320/1430 ... Exchange, 401 ... Radio base station B encoding method request, 402 ... Radio base station B encoding method response, 403 ... Switching between the code conversion system and the coding apparatus A, 404 ... Request for prohibiting the use of the coding apparatus B, 405 and 406 ... Setting of digital through mode, 407 and 408 ... Permission of use, 500 ... Voice code conversion unit, 510 ... code separation unit, 520 ... gain code fixed codebook code conversion unit, 530 ... code conversion control unit, 540
... Short-term prediction analysis parameter code conversion unit, 550 ... Power value code conversion unit, 560 ... Long-term prediction lag code conversion unit, 570
... code multiplexing unit, 580 ... transmission channel decoding unit, 590 ... transmission channel coding unit, 610 ... short-term prediction analysis parameter interpolation unit,
620 ... Short-term prediction analysis parameter code conversion control unit, 63
0 ... Short-term prediction analysis parameter code corresponding unit, 710 ... Power value interpolation unit, 720 ... Power code conversion control unit, 730 ... Power code corresponding unit, 810 ... Long-term prediction lag interpolation unit, 820 ... Long-term prediction lag code conversion control Unit, 830 ... Long-term predicted lag code corresponding unit, 1000 ... Input side excitation signal generation unit, 1001 ...
Adaptive codebook A, 1002 ... Fixed codebook A,
1003 ... Amplifier A, 1004 ... Amplifier B, 1005 ...
Gain codebook A, 1010 ... Output side excitation signal generation unit, 1011 ... Adaptive codebook B, 1012 ... Fixed codebook B, 1013 ... Amplifier C, 1014 ... Amplifier D, 1015 ... Gain codebook B, 1020 ... Error Minimum code selection unit, 1030 ... Error calculation unit, 1140.1
320 ... Multiplexing device A, 1150/1330 ... Multiplexing device B, 1212 ... Connection destination information, 1230/1300 ... Telephone A, 1240/1370 ... Telephone B, 1260/1
380 ... Telephone C, 1220/1340 ... Exchange C, 1
350 ... Multiplexing device C, 1400 ... Mobile terminal, 1410
… Adapter, 1500… Code conversion hardware, 1
510 ... DSP, 1520 ... Memory.

Claims (12)

[Claims] 1. A short code obtained by analyzing a voice signal by code excitation linear prediction at predetermined analysis time intervals, separating the spectrum envelope information and a spectrum fine structure, and quantizing and encoding the spectrum envelope information. A temporal analysis code, a power code obtained by quantizing and encoding the power of a voice signal, and the spectrum fine structure, and according to a predetermined prediction lag range of a long-term prediction lag filter, a filter coefficient of the long-term prediction lag filter. By the first speech coding method for coding the speech signal by multiplexing the long-term prediction lag code obtained by quantizing and coding the A coded linear predictive coding method, in which the coded multiplexed code is different from the first speech coding method in quantization value. Is a code conversion device for converting into a multiplex code according to a second audio encoding method, wherein the multiplex code encoded according to the first audio encoding method is input and separated for each code. And a code separating unit that separates the codes separated by the code separating unit according to the correspondence relationship between the code according to the first speech encoding method and the code according to the second speech encoding method. And a multiplexing unit that multiplexes each code by the second audio coding method converted by the conversion unit. Code converter. 2. The conversion unit according to claim 1, further comprising: a table showing the correspondence relationship between the code according to the first speech coding method and the code according to the second speech coding method. A code conversion device which performs the conversion by referring to the code conversion device. 3. The vector sum excitation linear predictive coding method (VSEL) according to claim 1, wherein the first speech coding method is
P: Vector-Sum Excited Linear Predictive Coding)
In the second speech coding method, the pitch-synchronous code excitation linear predictive coding method (PSI-CELP: Pitch Sync) is used.
hronous Innovation CELP), PSI-C
The ELP further includes a noise excitation code that quantizes and encodes a noise excitation component in the spectral fine structure, and the conversion unit converts the short-time analysis code encoded by the first speech encoding method. , A short-time analytic code according to the second speech coding method is converted into a short-time analytic code according to the second speech coding method according to a correspondence relationship between the short-time analytic code according to the first speech coding method and the short-time analytic code according to the second speech coding method. A short-term predictive analysis code conversion unit, a power code encoded by the first speech encoding method, a power code obtained by the first speech encoding method, and a power code obtained by the second speech encoding method. Power code conversion means for converting into a power code according to the second speech coding method, and a long-term prediction lag code coded according to the first speech coding method, According to the correspondence relationship between the long-term predicted lag code according to the first speech coding method and the long-term predicted lag code according to the second speech coding method, the long-term predicted lag code is converted into the long-term predicted lag code according to the second speech coding method. A long-term prediction lag code conversion means, and a long-term prediction lag code and a gain code encoded by the first speech coding method to generate an excitation signal in the first speech coding method, and the long-term prediction lag code. The excitation signal in the second speech encoding method is generated according to the long-term prediction lag code converted by the second speech encoding method by the conversion means, and the excitation signal in the first speech encoding method and the excitation signal in the first speech encoding method are generated. The gain code and the noise source code according to the second speech coding method that minimize the error from the excitation signal in the second speech coding method are output. And a noise excitation code conversion means. 4. The long-term prediction lag code converting means according to claim 3, wherein the long-term prediction lag codes of the first speech coding method and the second speech coding method are different from each other in the prediction lag range. To determine whether the long-term prediction lag code by the first speech coding method is within the prediction lag range of the second speech coding method, and determine the long-term prediction lag by the first speech coding method. If it is determined that the code is not within the prediction lag range of the second speech encoding method, then the second
The long-term predicted lag code of the second speech coding method is converted so as to be within the prediction lag range of the second speech coding method, and the long-term predicted lag code of the first speech coding method is converted into the second long-term predicted lag code.
If it is determined to be within the prediction lag range of the second speech coding method, the long-term prediction lag code of the first speech coding method is converted to the long-term prediction lag of the second speech coding method after conversion. A code conversion device having a code. 5. The long-term prediction lag code conversion means according to claim 4, wherein the long-term prediction lag code by the first speech coding method is determined not to be within the prediction lag range of the second speech coding method. If the long-term prediction lag code by the first speech coding method is smaller than the minimum value of the prediction lag range of the second speech coding method, the second speech coding method The long-term prediction lag code of the second speech coding method having a value larger than the minimum value of the prediction lag range and being N times the long-term prediction lag code of the first speech coding method (where N is A natural number of 2 or more) is used as the long-term predicted lag code by the second speech coding method after conversion, and the long-term predicted lag code by the first speech coding method is the second speech coding method. Must be within the predicted lag range of If it is determined that the long-term prediction lag code by the first speech coding method is larger than the maximum value of the prediction lag range of the second speech coding method, the immediately preceding converted first A code conversion device, wherein the long-term predicted lag code of the second speech coding method is used as a long-term predicted lag code after the conversion by the second speech coding method. 6. The second speech coding according to claim 1, wherein the conversion unit is configured to perform the second speech coding when the analysis time intervals of the first speech coding method and the second speech coding method are different from each other. A code conversion device, comprising: an interpolating means for interpolating a code by the second speech coding so as to output a code by the second speech coding at each analysis time interval of the method. 7. The short time prediction analysis code conversion means according to claim 3, wherein the first speech coding method and the second speech coding method have different short time analysis code analysis time intervals. When the analysis time interval of the short time analysis code of the first speech encoding method is shorter than the analysis time interval of the short time analysis code of the second speech encoding method, the second speech The short time analysis code of the second speech coding method after conversion is thinned out and output in accordance with the analysis time interval of the short time analysis code of the coding method, and the short time analysis of the first speech coding method is performed. When the code analysis time interval is longer than the analysis time interval of the short time analysis code of the second speech coding method, the short time analysis codes of the first speech coding method which are adjacent to each other in terms of analysis time are selected. Short time of the second speech coding method converted respectively From No. 析符, linearly seeking short analysis code of the second speech encoding method in accordance with the analysis time interval short analysis code of the second speech encoding method,
A code conversion comprising an interpolating means for outputting the obtained short time analysis code of the second speech encoding method in accordance with an analysis time interval of the short time analysis code of the second speech encoding method. apparatus. 8. A short code obtained by analyzing a voice signal by code excitation linear prediction at predetermined analysis time intervals, separating into spectral envelope information and spectral fine structure, and quantizing and encoding the spectral envelope information. A temporal analysis code, a power code obtained by quantizing and encoding the power of a voice signal, and the spectrum fine structure, and according to a predetermined prediction lag range of a long-term prediction lag filter, a filter coefficient of the long-term prediction lag filter. By the first speech coding method for coding the speech signal by multiplexing the long-term prediction lag code obtained by quantizing and coding the A first encoder for encoding, and a code excitation linear predictive coding method, wherein the quantization value is different from that of the first speech coding method A second decoder that performs decoding by a second speech encoding method that performs encoding, and a multiplexing code that is performed by the first encoder
The second code is converted into a multiplex code by the voice coding method of
And a code conversion device for outputting to the decoder, wherein the code conversion device inputs a multiplex code encoded by the first speech encoding method, A code separation unit that separates each code and each code separated by the code separation unit according to the correspondence relationship between the code according to the first speech coding method and the code according to the second speech coding method. A conversion unit for converting each code according to the second speech encoding method, and a multiplexing unit for multiplexing each code according to the second speech encoding method converted by the conversion unit. A code conversion system characterized by. 9. The code conversion device according to claim 8,
A connection destination information storage unit that stores in advance information indicating whether the connection destination encoding method is the first voice encoding method or the second voice encoding method, and the connection destination information storage unit. With reference to the above, when the encoding method of the connection destination is the second speech encoding method, the code conversion apparatus further comprises control means for controlling the code conversion apparatus to perform code conversion. system. 10. A short code obtained by analyzing a voice signal by code excitation linear prediction at each predetermined analysis time interval, separating the spectrum envelope information and a spectrum fine structure, and quantizing and encoding the spectrum envelope information. Time analysis code,
A power code obtained by quantizing and coding the power of a voice signal, and representing the spectral fine structure, according to a predetermined prediction lag range of the long-term prediction lag filter, the filter coefficient of the long-term prediction lag filter is quantized and coded. Encoded by the first speech encoding method for encoding the speech signal by multiplexing the encoded long-term prediction lag code and the gain code obtained by quantizing and encoding the amplitude of the spectral fine structure, A first communication device for receiving a signal output from a terminal; and a code excitation linear predictive coding method, wherein a coding value or a quantization method different from that of the first speech coding method is used. The second encoding is performed by the second audio encoding method, and the encoded signal is transmitted to the terminal.
And a third communication device that performs encoding by the first voice encoding method, the first communication device is encoded by the first voice encoding method. A code separation unit for inputting the multiplexed code and separating each code, and each code separated by the code separation unit, a code by the first speech encoding method and a second speech. A conversion unit that converts each code according to the second speech coding method into each code according to the correspondence with the code according to the coding method, and each code according to the second speech coding method converted by the conversion unit. A code conversion device including a multiplexing unit that multiplexes the communication device, an inquiry unit that inquires about an encoding method of a communication device that is a communication destination, and a communication device that is a communication destination in response to the inquiry by the inquiry unit When the encoding method of the communication apparatus of the communication destination is the second encoding method according to the response, the response is encoded by the first encoding method by the code conversion apparatus. Controlling the multiplex code to be converted into the multiplex code by the second speech coding method,
First communication means for notifying the communication apparatus of the communication destination that the second encoding method does not perform the encoding, and the second communication apparatus uses the inquiry means for encoding. When there is an inquiry about the encoding method, the second communication apparatus responds by saying that the encoding method in the second communication device is the second encoding method, and the control means of the case 1 from the second encoding. When there is a notification that encoding will not be performed by the method, the second information converted by the first communication device is not encoded by the second encoding method.
And a second control means for controlling the signal of the multiplex code by the voice coding method of (1) to be transmitted to the terminal. 11. A speech signal is analyzed by code excitation linear prediction at predetermined analysis time intervals, separated into spectral envelope information and spectral fine structure, and the spectral envelope information is quantized and encoded. Time analysis code,
A power code obtained by quantizing and coding the power of a voice signal, and representing the spectral fine structure, according to a predetermined prediction lag range of the long-term prediction lag filter, the filter coefficient of the long-term prediction lag filter is quantized and coded. In a communication device that encodes the voice signal by a first voice encoding method by multiplexing the encoded long-term prediction lag code and a gain code that is obtained by quantizing and encoding the amplitude of the spectral fine structure, A code conversion linear predictive coding method for converting into a multiplex code by a second speech coding method for performing coding with a quantization value or a quantization method different from that of the first speech coding method. The encoding device includes a code component for separating the multiplex code encoded by the first speech encoding method for each of the codes. Section and each of the codes separated by the code separation section according to the correspondence relationship between the code according to the first speech encoding method and the code according to the second speech encoding method. A communication device, comprising: a conversion unit that converts each code according to the encoding method, and a multiplexing unit that multiplexes each code according to the second voice encoding method converted by the conversion unit. 12. A speech signal is analyzed by code excitation linear prediction at predetermined analysis time intervals, separated into spectral envelope information and spectral fine structure, and the spectral envelope information is quantized and encoded. Time analysis code,
A power code obtained by quantizing and coding the power of a voice signal, and representing the spectral fine structure, according to a predetermined prediction lag range of the long-term prediction lag filter, the filter coefficient of the long-term prediction lag filter is quantized and coded. Coded the speech signal by multiplexing the encoded long-term prediction lag code, the gain code obtained by quantizing and encoding the amplitude of the spectral fine structure, and the noise excitation code obtained by quantizing and encoding the noise source component. A second excitation encoding method that encodes a multiplex code encoded by a first speech encoding method, which is a code excitation linear predictive encoding method and has a quantization value different from that of the first speech encoding method. A code conversion device for converting into a multiplex code according to the speech coding method of claim 1, wherein the multiplexed code coded according to the first speech coding method is input, And a code separation unit that separates each code, and a code that is separated by the code separation unit is associated with a code by the first speech encoding method and a code by the second speech encoding method. Therefore, it has a conversion unit that converts each code according to the second speech coding method, and a multiplexing unit that multiplexes each code according to the second speech coding method converted by the conversion unit. A code conversion device characterized by the above.