JPH05165497A - C0de exciting linear predictive enc0der and decoder - Google Patents

Abstract

PURPOSE:To provide the code exciting linear predictive encoder and decoder which can reproduce voice signals containing pulse signals with high quality, can easily deal with low encoding speed and can improve the quality of repro duced voices. CONSTITUTION:A pulse code book 22 storing pulse codes composed of isolate impulse is provided and used alternatively with a statistic code book 21. An LSP analysis part 11 obtains an LSP parameter from the input voice signal, and this LSP parameter is coded and transmitted by an LSP parameter coding part 12. A frequency characteristic operation part 28 operates the frequency characteristic of a statistic code or the pulse code so as to make it close to the frequency characteristic of the input voice signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a code-excited linear predictive coding (CELP) coder / decoder.

[0002]

2. Description of the Related Art Conventionally, as a high-efficiency coding system for voice signals (including acoustic signals) in the field of digital mobile communication,
A code-excited linear predictive coding system and a vector addition-excited linear predictive coding system (VSELP), which is a modification of the code-excited linear predictive coding system, have been adopted.

The basic structure of a coding system for a voice signal is to obtain a vocal tract parameter expressing a vocal tract characteristic of a voice and a sound source parameter expressing a source information. In the recent code-excited linear predictive coding method, the excitation signal as the sound source information is composed of an adaptive code that contributes to a voiced sound with statistically strong periodicity and a statistical code that contributes to a random unvoiced sound with statistically weak periodicity. Find the optimum adaptive code and statistical code in each codebook so that the sum of powers of the weighted errors between the input speech signal and the synthesized speech signal is minimized. The encoding process is performed in. At least a sound source parameter, whether it is a forward-type coding method that obtains a vocal tract parameter from an input speech signal or a backward-type coding method that obtains a vocal tract parameter from a synthesized speech signal,
Therefore, information of the optimum adaptive code and statistical code is transmitted.

It is known that such a code-excited linear predictive coding system can provide high-quality reproduced speech at a coding speed of 6 kbit / s to 8 kbit / s.

[0005]

However, some communication systems require a lower coding rate. For example, there is one that obtains a coding rate of 4 kbit / s or less. In the case of such a low speed, whether it is a forward type that transmits both vocal tract parameters and sound source parameters,
Whether it is a backward type that transmits sound source parameters,
Naturally, the number of coded bits assigned to the excitation parameter is small, and the number of adaptive codes and statistical codes stored in the adaptive codebook and the statistical codebook is also small. As a result, the quality of reproduced voice is deteriorated at such a low encoding speed.

Further, since the adaptive codebook is adaptively updated by the combined code of the optimal adaptive code and the statistical code, it can be said that the adaptive code is formed based on the statistical code. .. As a result, the voiced sound having a strong periodicity rises slowly, and a clear code having a strong pulse cannot be formed even in the stationary part of the voiced sound, so that the reproduced voice lacks intelligibility.

The present invention has been made in consideration of the above points, and a code capable of obtaining a reproduced voice of high quality even when a noise component having a strong pulse characteristic is included in the input voice signal. The present invention seeks to provide an excited linear predictive encoder and decoder.

Further, the present invention has been made in consideration of the above points, and even in the case of a low coding speed, the code excitation which can improve the quality of reproduced voice mainly from the sound source parameter side. The present invention seeks to provide a linear predictive encoder and decoder.

[0009]

In order to solve the above problems, according to the present invention of claim 1, an adaptive codebook is provided in a code excitation linear predictive encoder including an adaptive codebook and a statistical codebook as an excitation codebook. In addition to the statistical codebook, a pulsed codebook storing pulsed codes consisting of isolated impulses is provided.

In the code-excited linear predictive encoder of the present invention as defined in claim 2, the excitation code from the statistical codebook or the pulse-like codebook provided in this way is selected and used, and the selection information is selected. It was decided to send it to the code-excited linear predictive decoder.

A code-excited linear predictive encoder according to a third aspect of the present invention includes a pulse codebook, and the vocal tract parameters output to the code-excited linear predictive decoder are line spectrum pair parameters.

According to a fourth aspect of the code-excited linear predictive encoder of the present invention, the frequency characteristic operation for bringing the frequency characteristic of the fixed code outputted from the statistical codebook and / or the pulse codebook close to the frequency characteristic of the input speech signal. Section.

A code-excited linear predictive decoder according to the present invention of claim 5 corresponds to the encoder of claim 1,
In addition to the adaptive codebook and the statistical codebook, a pulsed codebook storing pulsed codes composed of isolated impulses is provided.

The code-excited linear predictive decoder according to the present invention of claim 6 uses the excitation code from the statistical codebook and the above-mentioned pulse-like codebook to correspond to the code-excited linear predictive coding according to the present invention of claim 2. It is characterized in that it is selected and used based on the selection information given from the container.

In the code-excited linear predictive decoder according to the present invention as defined in claim 7, the vocal tract parameter given from the corresponding code-excited linear predictive encoder is a line spectrum pair parameter, which is utilized for speech reproduction. Is characterized by.

The code-excited linear predictive decoder according to the present invention of claim 8 corresponds to the frequency characteristic of the fixed code outputted from the statistical codebook and / or the pulse-like codebook, and corresponds to the code excitation of the present invention according to claim 4. It is characterized in that it has a frequency characteristic operation section that operates based on information given from the linear predictive encoder.

[0017]

According to the code-excited linear predictive encoder of the present invention as defined in claim 1 and the code-excited linear predictive decoder as claimed in claim 5 of the present invention, it is possible to speed up the rise of a voiced sound having a strong periodicity, and In addition to the adaptive codebook and statistical codebook, a pulsed codebook that stores pulsed codes consisting of isolated impulses is provided so that a strong excitation signal with a clear pulsed characteristic can be formed even in the stationary part of Is.

According to the code-excited linear predictive encoder of the present invention as defined in claim 2, a statistical codebook or a pulse codebook is provided so that a low coding rate can be realized even when a pulse codebook is provided. It was decided to use the excitation code from. Therefore, the code-excited linear predictive decoder according to the present invention of claim 6 corresponding to this encoder also
The excitation code from the statistical codebook or the pulse codebook is selected and used. Selection information required at this time is given from the encoder.

A code-excited linear predictive encoder according to a third aspect of the present invention relates to a code-excited linear predictive encoder provided with a pulse codebook for improving reproduction quality at a low coding rate from the viewpoint of excitation parameter. In consideration of the vocal tract parameter in terms of the improvement of the playback quality in, the vocal tract parameter given to the code-excited linear predictive decoder is defined as the line spectrum pair parameter. Therefore, the code-excited linear predictive decoder according to the present invention of claim 7 corresponding to this encoder also utilizes this line spectrum pair parameter for speech reproduction.

The claims 3 and 7 are limited to a so-called forward type code excitation linear predictive encoder and decoder.

In the code-excited linear predictive encoder of the present invention as defined in claim 4, the frequency characteristic manipulating section determines the frequency characteristic of the fixed code outputted from the statistical codebook and / or the pulse codebook from the frequency of the input speech signal. Perform an operation that approaches the characteristics. This is because the frequency characteristic of the excitation signal has been theoretically modeled as white in the past, but in reality it is not white and has a characteristic close to the frequency characteristic of the input audio signal. Is experimentally confirmed, and if the frequency characteristics of the statistical code and the pulse-like code are brought close to the frequency characteristics of the input speech signal, it is possible to obtain a high quality synthesized speech signal. This is because the effective frequency component of the excitation signal is considerably larger than the quantization error signal and the masking effect of the quantization error signal is obtained. Therefore, claim 8 corresponding to this encoder
The code-excited linear predictive decoder of the present invention also has a frequency characteristic operation unit for performing similar processing.

[0022]

【Example】

(A) One Embodiment of Code Excited Linear Predictive Encoder One embodiment of the code excited linear predictive encoder according to the present invention will be described in detail below with reference to the drawings. Here, FIG. 1 is a block diagram showing the configuration of this embodiment.

In FIG. 1, the code-excited linear predictive encoder of this embodiment is mainly composed of an input speech processing unit 1, an optimum synthesized speech searching unit 2 and a multiplexing unit 3.

The input speech processing unit 1 includes an LSP (line spectrum pair parameter) analysis unit 11, an LSP parameter coding unit 12, an LSP parameter decoding unit 13, an LPC (linear prediction coefficient) coefficient conversion unit 14, and a weighting filter 15. , A synthesis filter zero-input response generation unit 16, a weighting filter zero-input response generation unit 17, and two subtractors 18 and 19, and transmits them to a decoder when an input voice signal is given. The vocal tract parameters are obtained and the target voice signal of the synthesized voice signal formed by local reproduction is formed.

In the case of this embodiment, the digitized discrete input speech signal sequence is accumulated for a time corresponding to the analysis frame length for obtaining the vocal tract parameter, and further, this analysis frame length is several. It is divided into subframes and processed by the input voice processing unit 1.

The input voice signal is given to the LSP analysis section 11, and is subjected to LSP analysis by this LSP analysis section 11 and converted into an LSP parameter as a vocal tract parameter.
This LSP parameter is the LSP parameter encoding unit 12
Is encoded (for example, vector quantization), given to the multiplexing unit 3, and transmitted to the code excitation linear prediction decoder side. Also, the encoded LSP parameters are LS
After being decoded (vector inverse quantization) by the P parameter decoding unit 13, the LPC coefficient conversion unit 14 outputs LP.
Converted to C coefficient. The LPC coefficients converted in this way are used as tap coefficients for the weighting filter 15, the synthesis filter zero-input response generation unit 16, the weighting filter zero-input response generation unit 17, and the weighted synthesis filter 29 described later. It is also given to the frequency characteristic operation unit 28 described later. In addition, the LSP parameter output from the LSP analysis unit 11 is not directly converted into an LPC coefficient, but a code,
The LSP parameters that have been subjected to the decoding process are converted into LPC coefficients by the same LP as the LPC coefficients used by the decoder.
This is because the sound source parameter can be appropriately determined by using the C coefficient in local reproduction.

Here, the LSP parameter is used as the vocal tract parameter for transmission because the interpolation characteristic with respect to the frequency characteristic of the vocal tract is improved, and the LSP parameter is LPC even if it is encoded with a small number of encoding bits. This is because the distortion given to the vocal tract spectrum is smaller than the parameters, etc., and efficient coding can be performed in combination with the vector quantization method.

Next, the operation of forming the target voice signal for the locally reproduced synthesized voice signal from the input voice signal will be described.

The above-mentioned input audio signal is a weighting filter 1
5 and weighted in consideration of human auditory characteristics, and then given to the subtractor 18 as a subtracted input.
The subtracter 18 is also supplied with a zero input response signal regarding the weighted synthesis filter 29 generated by the synthesis filter zero input response generation unit 16 using the LPC coefficient as a tap coefficient, as a subtraction input. Thus, an audio signal in which the influence of the state of the weighted synthesis filter 29 in the immediately preceding analysis frame is removed is obtained, and this audio signal is given to the subtractor 19 as a subtracted input. The weighting filter 1 generated by the weighting filter zero input response generation unit 17 using the LPC coefficient as the tap coefficient is also included in the subtractor 19.
The quiescent response signal for 5 is provided as the subtraction input. Thus, an audio signal in which the influence of the state of the weighting filter 15 in the immediately preceding analysis frame is removed is obtained,
This is given to the subtractor 30 which will be described later as a target voice signal.

The optimum synthesized speech search unit 2 searches for a sound source parameter whose synthesized speech signal produced by local reproduction is most similar to the target speech signal. The adaptive codebook 20, statistical codebook 21, pulse codebook 22, Gain codebook 23, gain controllers 24 and 27, adder 2
5, fixed code selection switch 26, frequency characteristic operation unit 2
8, a weighted synthesis filter 29, a subtractor 30, an error power sum calculation unit 31, and a minimum error power sum code selection unit 32.

The adaptive codebook 20, the statistical codebook 21, and the pulse codebook 22 store an adaptive code, a statistical code, and a pulse code, which are waveform codes related to the excitation signal, respectively. Reference numeral 23 stores a gain code relating to the adaptive code and the fixed code (the statistical code and the pulse characteristic code are collectively referred to as such).

The adaptive code and the statistical code are a waveform code contributing to a voiced sound having a statistically strong periodicity and a waveform code contributing to a random unvoiced sound having a statistically weak periodicity, as in the conventional case. In addition, adaptive codebook 2
The adaptive code of 0 is adaptively updated as described later.
The pulse code is a waveform code composed of isolated impulses. The pulse code considers that it contributes to the rise of a voiced sound with a strong periodicity and the steady part of a voiced sound with a clear pulse. The gain code is, for example, vector-quantized, and one component of the vector relates to the gain of the adaptive code and the other component relates to the gain of the fixed code.

The pulsed sound source signal is a simple signal having periodicity and may be generated by the pulse signal generator. However, as in this embodiment, it is coded and read from the codebook 22. This is preferable because of the following reasons. That is, it is easy to synchronize with the output from the adaptive codebook 20, and by adopting the same book configuration as the statistical codebook 21, the statistical code or the pulse code is selected and transmitted to the decoder as described later. This is because the multiplexing process and so on are facilitated.

A code-excited linear predictive decoding is performed by finding an optimum code of various codes in which the synthesized speech signal locally reproduced using such various codes is most similar to the target speech signal and giving the index to the multiplexing section 3. It is transmitted to the chemical converter side. In this embodiment, since a low coding rate is taken into consideration, a statistical code or a pulsed code is selected for a fixed code and its index is transmitted. Therefore, the selection information indicating which one is selected as the fixed code is also transmitted to the code excitation linear prediction decoder side.

In the case of this embodiment, the search for the optimum code (including the selection process of the statistical code or the pulsed code) is executed in the order of the adaptive code, the statistical code, the pulsed code, and the gain code.

At the time of searching for the optimum adaptive code, the outputs from the statistical codebook 21 and the pulse codebook 22 are set to 0, and the gain controller 24 multiplies an appropriate value of the gain coefficient (for example, 1). It is done like this. In such a state, all the adaptive codes stored in the adaptive codebook 20 are output in time sequence or in parallel, and are output as an excitation signal to the weighted synthesis filter 29 via the gain controller 24 and the adder 25. give. The weighted synthesis filter 29 performs convolution processing on this excitation signal using the LPC coefficient provided from the LPC coefficient conversion unit 14 as a tap coefficient, and outputs a synthesized speech signal in which only the content of the adaptive code is reflected as a sound source parameter. Ask for all adaptive codes.

The subtractor 30 obtains an error signal between the synthesized voice signal and the target voice signal, in which only the content of the adaptive code is reflected, for all the adaptive codes, and supplies it to the error power sum calculation unit 31. The error power sum calculation unit 31 obtains the sum of squares of the components of the error signal (error power sum) for all adaptive codes and supplies the sum to the minimum error power sum code selection unit 32.
The minimum error power code selection unit 32 determines the adaptive code with the minimum error power sum as the optimum one.

Next, the search for the optimum statistical code is executed. At the time of this search, the fixed code selection switch 26 is switched to the statistical codebook 21 side, and the adaptive codebook 20 sets the output to 0 ( It is also possible to output the optimal adaptation code). In such a state, all statistical codes stored in the statistical code book 21 are time-sequentially or in parallel output, and input to the frequency characteristic operation unit 28 via the fixed code selection switch 26 and the gain controller 24. Let

The frequency characteristic operation unit 28 has a detailed configuration shown in FIG. 3 which will be described later, and changes the frequency characteristic of the input statistical code into the frequency characteristic of the input voice signal corresponding to the temporal length of the statistical code. The conversion operation is performed so as to approach. In this way, all the statistical codes whose frequency characteristics have been converted are supplied to the weighted synthesis filter 29 as excitation signals via the adder 25 (which is equal to the value in this case). After that, the process is performed in the same manner as the search for the optimum adaptive code, and the minimum error power sum code selection unit 32 determines the optimum statistical code.

The frequency characteristic operating section 28 is provided for the following reason. Conventionally, the frequency characteristic of the excitation signal has been theoretically modeled as white, but it has been experimentally confirmed that the frequency characteristic of the excitation signal is not white and has characteristics close to those of the input audio signal. There is. Therefore, if the frequency characteristics of the statistical code and the pulse-like code are brought close to the frequency characteristics of the input speech signal, a higher quality synthesized speech signal can be obtained, and the effective frequency component of the excitation signal is quantized. The masking effect of the quantized error signal can be obtained by making it considerably larger than the error signal. Therefore, the frequency characteristic operation unit 28 is provided. Here, as the information representing the frequency characteristic of the input speech signal, there is the LPC coefficient, and there is also the information (including the gain for it) of the optimum adaptive code that means the pitch prediction information. Therefore, the frequency characteristic operation unit 28 operates the frequency characteristics of the statistical code and the pulse code based on these pieces of information.

When the search for the optimum statistical code is completed in this way, the search for the optimum pulse code is next performed. During this search, the fixed code selection switch 26 is switched to the pulse codebook 22 side, and the adaptive codebook 20 sets the output to 0 (note that the optimal adaptive code may be output). In such a state, all the pulsed codes stored in the pulsed code book 22 are output in time sequence or in parallel. Subsequent processing is the same as that at the time of searching for the optimum statistical code, and therefore its explanation is omitted.

When the optimum pulse-like code is determined in this way, the minimum error power sum code selecting unit 32
Compares the error power sum of the optimum statistical code and the error power sum of the optimum pulsed code, and determines the smaller error power sum as the fixed code to be transmitted to the code excitation linear prediction decoder side.

After that, the optimum gain code is searched. At the time of searching for this gain code, the adaptive codebook 20 outputs the optimum adaptive code, and the fixed code selection switch 26 selects the statistical codebook 21.
Alternatively, the pulse code code book 22 is switched to and the optimum fixed code is output from the selected fixed code book 21 or 22. One gain code book 23 is composed of a gain for adaptive code and a gain for fixed code. The gain for adaptive code is given to the gain controller 24 and the gain for fixed code is given to the gain controller 27. Thus, the optimum adaptive code whose gain has been controlled and the optimum fixed code which has been subjected to the frequency characteristic operation and gain control are added by the adder 25, and are added to the weighted synthesis filter 29 as an excitation signal. Such processing is performed by the gain codebook 23.
For all gain codes in time sequence or in parallel. The processing at the time of searching after the weighted synthesis filter 29 is the same as the processing at the time of searching for other codes.

When the optimum adaptive code, the optimum fixed code, and the optimum gain code are obtained, the minimum error power sum code selecting section 32 gives these indexes to the multiplexing section 3 and selects either the statistical code or the pulse code. The fixed code selection switch information indicating whether the selection has been made is also given to the multiplexing unit 3. The multiplexing unit 3 multiplexes the LSP parameter given from the LSP parameter coding unit 12 and these pieces of information, and outputs the multiplexed information to the code excitation linear prediction decoder side. When vector quantization is applied as the gain code, the transmitted index is the vector number.

In addition, the minimum error power sum code selection unit 32
Of the index and fixed code selection switch information to be given to the multiplexing unit 3 in the corresponding codebooks (20 and 2).
3 and 21 or 22) and the fixed code selection switch 26. At this time, the switch 26 is switched, and the optimum code is output from each codebook. As a result, an excitation signal that can form a synthesized speech signal that is closest to the target speech signal during the current subframe processing is output from the adder 25, and this is provided to the adaptive codebook 20.
Then, the adaptive code book 20 performs an adaptive code update process.

The above encoding process is repeated for each subframe, and encoded speech signals are sequentially transmitted to the code excitation linear predictive decoder.

FIG. 3 shows a detailed configuration of the above-mentioned frequency characteristic operating unit 28. In FIG. 3, the frequency characteristic operating unit 28 includes two filters 28 connected in cascade.
1 and 282 and a pitch lag determining unit 283.

The fixed code output from the fixed code selection switch 26 is applied to the first filter 281.
Impulse response H2 (z) of this first filter 281
Is selected as shown in the following equation, and the frequency conversion operation is performed on the fixed code input by this.

H1 (z) = (1−Σγ ⁱ ai z ⁻ⁱ ) / (1−Σ v ⁱ ai z ⁻ⁱ ) ... (1) where ai (i is 1 to M) is the LPC coefficient conversion It is a tap coefficient for the synthesis filter 29 supplied from the unit 14. Further, γ and ν are constants smaller than 1 which are set in advance.

The fixed code of which the frequency characteristic is manipulated by the first filter 281 is the second code 282.
Entered in. The pitch lag determining unit 283 obtains the pitch lag L from the index of the optimum adaptive code for the adaptive codebook 20 and supplies it to the second filter 282. The impulse response H2 (z) of this second filter 282 is
It is selected as shown in the following equation, and the frequency conversion operation is performed on the fixed code input by this.

H2 (z) = 1 / (1-εz ^−L ) (2) where ε is a predetermined constant smaller than 1. The output of the second filter 282 is given to the gain controller 27.

As described above, the frequency characteristic manipulating unit 28 having such a detailed configuration converts the frequency characteristic of the input fixed code into the frequency characteristic of the input voice signal corresponding to the temporal length of the fixed code. It is designed to be close.

Therefore, according to the code-excited linear predictive encoder of the above embodiment, it is possible to obtain a reproduced voice of high quality even at a low encoding speed. Hereinafter, it will be specifically described that such an effect is obtained.

(1) When a low coding rate is achieved, the number of coding bits assigned to excitation parameters is small, so the number of fixed codes prepared is small, and the pulse noise included in the input speech signal is clarified. However, in the case of this embodiment, since the pulse code is used, the reproduction quality of the voice in such a case can be improved.

Since the pulse code and the statistic code are switched and used, it is possible to cope with a low coding rate and to improve the reproduction quality for a signal in which a random signal and a pulse signal are mixed such as a voice transient portion. Can be increased.

(2) When a low coding rate is achieved, the number of coded bits for excitation parameters is reduced, but the number of coded bits for vocal tract parameters is also reduced. In the case of this embodiment, even if encoding is performed with a small number of encoding bits, the LP
Since the LSP parameter that causes less distortion in the vocal tract spectrum than C or the like is transmitted, the reproduction quality can be improved from this point.

(3) As described above, since the frequency characteristic operation unit is provided in consideration of the fact that the actual excitation signal has the frequency characteristic close to the frequency characteristic of the input voice signal, it is appropriate for the actual use. It is possible to improve the playback quality only,
Along with this conversion, a masking effect for the quantized error signal is generated, and the reproduction quality can be improved.

(B) One Embodiment of Code-Excited Linear Prediction Decoder Next, one embodiment of the code-excited linear prediction decoder according to the present invention will be described in detail with reference to the drawings. This embodiment corresponds to the embodiment of the code excitation linear predictive encoder shown in FIG. FIG. 2 is a block diagram showing the configuration of this embodiment.

In FIG. 2, the code-excited linear predictive decoder according to this embodiment has a demultiplexing section 40, an LSP parameter decoding section 41, an LPC coefficient conversion section 42, an adaptive codebook 43, a statistical codebook 44, and pulse characteristics. Codebook 45, gain codebook 46, gain controllers 47, 49,
Fixed code selection switch 48, frequency characteristic operation unit 50,
It is composed of an adder 51 and a weighted synthesis filter 52.

The coded speech signal supplied from the code excitation linear predictive encoder side is input to the demultiplexing unit 40.
The demultiplexing unit 40 demultiplexes the encoded voice signal into LSP parameters, an index of the optimum adaptive code, an index of the optimum fixed code, an index of the optimum gain code, and fixed code selection switch information. And LS
The P parameter is given to the LSP parameter decoding unit 41,
The index of the optimum adaptive code is the adaptive code book 43.
To the gain code book 46 and the fixed code selection switch information to the fixed code selection switch 48. In addition, the index of the optimum fixed code is given to the statistical codebook 44 or the pulse codebook 45 determined based on the fixed code selection switch information.

The LSP parameter decoding unit 41 decodes (for example, vector dequantization) the given encoded LSP parameter, and the LPC coefficient conversion unit 42 converts this LSP parameter into an LPC coefficient. The LPC coefficient converted in this way is given to the frequency characteristic operation unit 50 and the weighted synthesis filter 52 as vocal tract parameter information.

The gain code book 46 separates a gain code determined by the given index into an adaptive code gain code and a fixed code gain code, and outputs them to an adaptive code gain controller 47 and a fixed code gain controller 49, respectively. give.

The adaptive code book 43 outputs an adaptive code determined by the given index, and this adaptive code is gain-controlled by the gain controller 47 and is given to the adder 51. Further, the adaptive code book 43 provides the adaptive code to the frequency characteristic operation unit 50.

The statistical codebook 44 or the pulse codebook 45 stores the statistical code or pulse code corresponding to the given index in the fixed code selection switch 4.
8 is applied to the frequency characteristic operation unit 50, and the frequency characteristic operation unit 50 operates the frequency characteristic based on the LPC coefficient and the index of the adaptive code (the operation is performed so as to be close to the frequency characteristic of the input voice signal). ). The detailed configuration of the frequency characteristic operation unit 50 has the configuration shown in FIG. 3 described above. The fixed code whose frequency characteristic is manipulated in this way is gain-controlled by the gain controller 49 and is given to the adder 51.

The adder 51 adds the given adaptive code and fixed code and gives the addition signal as an excitation signal to the weighted synthesis filter 52. The weighted synthesis filter 52 convolves the excitation signal with the LPC coefficient to obtain and output a synthesized speech signal.

The excitation signal output from the adder 51 is also supplied to the adaptive codebook 43. At this time, the adaptive code book 43 updates the adaptive code using this excitation signal.

The code-excited linear predictive decoder performs the above-described processing every time a coded speech signal is given, that is, every subframe.

Therefore, according to the code-excited linear predictive decoder of this embodiment, it has a configuration for processing the given LSP parameter, and the pulse-like codebook 45 is used as the sound source.
And the frequency characteristic manipulating unit 50 that brings the frequency characteristic of the fixed sound source closer to the frequency characteristic of the input speech signal. Therefore, the effect of the above-described code excitation linear predictive encoder becomes actual.

(C) Other Embodiments The features of the above embodiment are that the LSP parameter is used as the vocal tract parameter for transmission, and that the pulse codebook is provided as the source parameter. The point is that the frequency characteristics of the fixed code are manipulated. Of these, the characteristic configurations of the second point and the third point do not exist conventionally, and each may be independently incorporated in the encoder and the decoder.

Although the above embodiment relates to the forward type code excitation linear predictive encoder and decoder, the present invention is applied to the backward type code excitation linear predictive encoder and decoder. It can also be applied.

Further, although the above embodiment is constructed in consideration of the coding rate of 4 bits / s or less, it is needless to say that it can be applied to the encoder and the decoder of the coding rate higher than this. is there. If the coding speed permits, the statistical codebook and the pulse codebook may not be selective but may always be effective.

Frequency characteristic operating units 28 and 5 of the above embodiment
Although 0 performs both the frequency characteristic operation using the LPC coefficient and the frequency characteristic operation using the pitch lag, at least one operation may be performed.

[0073]

According to the code-excited linear predictive encoder and decoder of claims 1 and 5, a pulse-like code consisting of isolated impulses is stored in addition to the adaptive codebook and the statistical codebook. Since a pulse codebook is provided, it is possible to obtain a clear reproduced voice by speeding up the rise of voiced sound with strong periodicity and forming a strong pulsed excitation signal even in the stationary part of voiced sound. You can Further, it is possible to obtain a good reproduced voice even for a signal in a transitional period of a voice in which a random signal and a pulse-like signal are mixed.

According to the code-excited linear predictive encoder and decoder of claims 2 and 6, since the statistical code and the pulse-like code are selected and used, the number of encoded bits of the excitation parameter is small. It is possible to obtain good reproduced sound.

According to the code-excited linear predictive encoder and decoder of claims 3 and 7, since the vocal tract parameter used for speech synthesis is the line spectrum pair parameter, the vocal tract parameter is also taken into consideration. High quality at low coding speed can be realized.

According to the code-excited linear predictive encoder and decoder of claims 4 and 8, the frequency characteristic of the fixed code output from the statistical codebook and / or the pulse-like codebook is converted into the input speech signal. Since the operation of approaching the frequency characteristic is performed, the sound source parameters correspond well to the actual sound source, and the quality of the reproduced sound can be improved from this point.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a code-excited linear predictive encoder.

FIG. 2 is a block diagram showing an embodiment of a code-excited linear predictive decoder.

FIG. 3 is a block diagram showing a detailed configuration of a frequency characteristic operation unit 28 (50) of the embodiment.

[Explanation of symbols]

3 ... Multiplexing unit, 11 ... LSP parameter analyzing unit, 12 ...
LSP parameter coding unit, 13, 41 ... LSP parameter decoding unit, 14, 42 ... LPC coefficient conversion unit, 20,
43 ... Adaptive codebook, 21, 44 ... Statistical codebook, 22, 45 ... Pulse codebook, 25, 51 ...
Adder, 26, 48 ... Fixed code selection switch, 28,
50 ... Frequency characteristic operation unit, 29, 52 ... Weighted synthesis filter, 30 ... Subtractor, 31 ... Error power sum calculation unit, 32
... minimum error power code selection unit, 40 ... demultiplexing unit.

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Yoshihiro Ariyama 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd.

Claims (8)

[Claims] 1. A code-excited linear predictive encoder having an adaptive codebook and a statistical codebook as an excitation codebook, in which, in addition to the adaptive codebook and the statistical codebook, a pulsed code consisting of isolated impulses is stored. A code-excited linear predictive encoder having a pulse codebook. 2. The excitation code from the statistical codebook or the pulse codebook is selected and used, and the selection information is sent to a code excitation linear predictive decoder. Code-excited linear predictive encoder. 3. The code-excited linear predictive encoder according to claim 1, wherein the vocal tract parameters output to the code-excited linear predictive decoder are line spectrum pair parameters. 4. A code-excited linear predictive coding having a frequency characteristic operation unit for approximating the frequency characteristic of the excitation code output from the statistical codebook and / or the pulse codebook to the frequency characteristic of the input speech signal. vessel. 5. A code-excited linear predictive decoder having an adaptive codebook and a statistical codebook as an excitation codebook, wherein a pulsed code consisting of isolated impulses is stored in addition to the adaptive codebook and the statistical codebook. A code-excited linear predictive decoder having a pulse codebook. 6. The excitation code from the statistical codebook or the pulse codebook is selected based on selection information provided from a corresponding code excitation linear predictive encoder.
The code-excited linear predictive decoder according to claim 5, which is selected and used. 7. The code excitation according to claim 5, wherein the vocal tract parameters given by the corresponding code excitation linear predictive encoder are line spectrum pair parameters, which are used for speech reproduction. Linear predictive decoder. 8. A frequency characteristic manipulating section for manipulating the frequency characteristic of the fixed code output from the statistical codebook and / or the pulse codebook based on the information given from the corresponding code excitation linear predictive encoder. A code-excited linear predictive decoder having the above characteristics.