JP3416331B2 - Audio decoding device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech decoding apparatus for interpolating a speech signal with respect to a frame in which a coding error is detected on the decoding side in a mobile communication system or the like in which a coding error occurs on a transmission line.

[0002]

2. Description of the Related Art In recent years, the practical application of digital mobile communication systems has been rapidly advanced. In a mobile communication system, when a code error occurs on a transmission path and an error that cannot be corrected by error correction is detected, frame interpolation is often used in which voice decoding is performed using past voice encoding parameters.

FIG. 4 shows the structure of a conventional speech decoding apparatus of this type. In FIG. 4, 1 is transmitted from the encoder, separates the received data representing the coding information for each frame, which is a fixed section of speech, into respective speech coding parameters, and determines whether the frame is an interpolated frame. Parameter separation / interpolation determiner, 2 is a previous frame coding parameter holder that holds the coding parameter used for decoding in the previous frame, and 3 is a switch that switches the coding parameter used in that frame depending on whether it is an interpolation frame Units 4 and 4 are voice decoders that perform voice decoding of the frame using the coding parameters and frame interpolation information selected by the switch 3.

The operation of the speech decoding apparatus configured as above will be described below. First, the parameter separation / interpolation determiner 1 receives the transmission data transmitted from the encoder, and outputs the encoding parameter of the frame and frame interpolation information indicating whether the frame is an interpolation frame. When the frame is not an interpolated frame, the switch 3 is set to the side a, and the audio decoder 4 performs normal audio decoding using the encoded data of the frame. On the other hand, when the frame is an interpolated frame, the switch 3 is set to the b side, and speech decoding is performed using the encoding parameter of the previous frame stored in the previous frame encoding parameter holder 2.

[0005]

However, in the above conventional speech decoding apparatus, since the decoding parameters of the previous frame are used as they are in the interpolation frame for decoding, there is a deterioration in speech quality in the decoded speech of the interpolation frame. It had a problem that it occurred.

Specifically, CELP (Code Excited Linear Pred), which is a mainstream method for low bit rate speech coding, is used.
When generating an adaptive codebook sound source using the parameter (lag) that represents the pitch period of the previous frame in the interpolation method, the lag may not be the optimum value for the interpolation frame. An adaptive codebook sound source having different pitch periodicity may not be generated.

Further, in the case where the excitation gain parameter is encoded depending on the generated adaptive codebook excitation, if the excitation gain parameter of the previous frame is used as it is, the encoder and decoding of the adaptive codebook are performed. It is impossible to generate a driving sound source with an appropriate gain due to mismatch in the power supply.

Further, when the encoder has a structure having coding modes of voiced mode and unvoiced mode, conventionally, decoding is performed with the same mode and the same coding parameter as the mode of the previous frame of the interpolated frame. However, there are cases where voicedness is high even in the unvoiced mode, and conversely, voicedness is high even in the voiced mode, in which case the quality of the decoded speech of the interpolated frame deteriorates.

An object of the present invention is to solve the above-mentioned conventional problems, and an object thereof is to provide an excellent speech decoding apparatus capable of suppressing perceptual deterioration in decoded speech of an interpolated frame.

[0010]

In order to achieve the above object, the present invention has, as a first configuration, received data transmitted from an encoder and representing encoded information for each frame which is a fixed section of speech. as well as separated into the sign-parameter, and determines the parameter separation and interpolation determination unit the frame is whether the interpolation frame, and the frame coding parameter holder before holding the encoding parameters used for decoding the previous frame , A driving sound source holder that holds a driving sound source signal in a certain section of the sound decoded up to the preceding frame, and a plurality of driving sound source signals from the driving sound source holder and the preceding frame from the preceding frame coding parameter holder. using the pitch period parameter, the optimal peak from a plurality of pitch period parameter of the previous frame, the frame interpolation of the frame By using the interpolation pitch period determiner for determining a switch cycle parameters, and switch for switching the coding parameters used in the frame according to whether the interpolation frame, the coding parameter and frame interpolation information selected by said switcher It has a configuration including a voice decoder for performing voice decoding of the frame.

As a second configuration, the received data transmitted from the encoder and representing the coding information for each frame, which is a fixed section of speech, is separated into respective speech coding parameters, and the frame is interpolated frame. A parameter separation / interpolation determiner that determines whether or not, a current frame encoding parameter determiner that determines an encoding parameter used for voice decoding of the frame from frame interpolation information, and an adaptive codebook that stores past driving sound sources, A fixed codebook that holds a fixed source vector such as a noise source, an adaptive codebook during frame interpolation, and a source gain controller that controls those gains using signals generated from the fixed codebook, an adaptive codebook, and A multiplier that multiplies the signal generated from the fixed codebook by the gain and the driving sound It has a configuration in which an adder for generating a.

Further, as a third configuration, in a speech decoding apparatus having a coding mode corresponding to each of a voiced section and a unvoiced section of speech, the coding is performed for each frame which is a fixed section of speech transmitted from an encoder. A parameter separation / interpolation determiner for determining whether or not the frame is an interpolated frame, while separating the reception data representing the information into respective audio encoding parameters and encoding mode information,
A previous frame coding parameter holder that holds the coding parameter / coding mode used for decoding in the previous frame,
A driving excitation cage holding the excitation signal of a predetermined section of the audio decoded in up to the previous frame, when the excitation signal and the previous frame output from the driving excitation cage voiced encoding mode of the previous frame Multiple pitch period parameters
Use a meter when setting the frame interpolation
Determine the coding mode to be
In the voice coding mode, the multiple frames of the previous frame are
From among the pitch period parameters, the optimum pitch period parameter
The interpolation pitch cycle determiner that determines the input data, the mode information switch that switches the input of the encoding mode information at the time of decoding the frame according to the frame interpolation information, and the encoding parameter used in the frame depending on whether it is an interpolation frame or not. A coding parameter switch for switching, a mode switch for switching the coding mode according to the mode information obtained from the output of the mode information switch, and a coding parameter and frame interpolation information selected by the coding parameter switch are used. In this configuration, a voiced mode decoder and an unvoiced mode decoder for performing voice decoding of the frame are provided.

[0013]

According to the first aspect of the present invention, the driving sound source having more appropriate pitch periodicity is selected by selecting the optimum pitch period parameter used in frame interpolation from the plurality of pitch period parameters of the previous frame. Can be generated.

With the second configuration, the adaptive codebook sound source generated from the past driving sound source and having the correlation with the past driving sound source, and the fixed codebook having no correlation with the past driving sound source such as the noise component. The power ratio with the sound source is controlled in a fixed ratio or with a variable ratio according to the correlation of the past driving sound source to control the gain to generate the driving sound source. Voice decoding can be performed.

Further, according to the third configuration, the interpolated pitch period determiner evaluates the voicedness / unvoicedness of the previous frame based on the correlation of the past driving sound sources, and when the voicedness is high, the voiced mode is used. If not, by performing the decoding process of the interpolation frame in the unvoiced mode, it is possible to perform the voice decoding while suppressing the deterioration of the auditory sense.

[0016]

【Example】

(Embodiment 1) Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows the configuration of a speech decoding apparatus according to the first embodiment of the present invention. In FIG. 1, reference numeral 101 separates the received data, which is transmitted from the encoder and represents the coding information for each fixed section (frame) of the speech, into respective speech coding parameters and determines whether the frame is an interpolated frame. Parameter separation / interpolation determiner, 102 is a previous-frame coding parameter holder that holds the coding parameter used for decoding in the previous frame, and 103 holds the driving excitation signal of a certain section of the speech decoded up to the previous frame Drive sound source holder,
Reference numeral 04 indicates a plurality of pitch periods of the preceding frame from the driving sound source holder 103 and the preceding frame coding parameter holder or a parameter similar thereto,
An interpolation pitch cycle determiner for determining an optimum pitch cycle for the frame interpolation of the frame or a parameter similar thereto, 105 is a switch for switching an encoding parameter used in the frame depending on whether or not the frame is an interpolation frame, 106
Is a speech decoder that performs speech decoding of the frame using the coding parameter and the frame interpolation information selected by the switch 105.

The operation of the speech decoding apparatus configured as described above will be described below. First, the parameter separation / interpolation determiner 101 receives the transmission data transmitted from the encoder, separates it into the encoding parameters of the frame, determines whether the frame is an interpolation frame, and determines the encoding parameters and Output frame interpolation information. Here, the interpolated frame is a frame in which received data of the frame is missing or an error due to a transmission path error is detected, and is a frame in which clear decoding of the received data causes clear voice quality deterioration. If the frame is not an interpolated frame, the switch 105 is set to the a side, and the audio decoder 106 performs normal audio decoding using the encoded data of the frame.

On the other hand, when the frame is an interpolated frame, first, a plurality of pitch periods of the previous frame from the driving excitation and the previous frame coding parameter holder 102, which are the outputs of the driving excitation holder 103, or a parameter similar thereto is used. Then, by the interpolation pitch period determiner 104,
The optimum pitch period for frame interpolation of the frame or a similar parameter is determined.

An example of the determination method is shown below. First, the pitch period obtained by decoding the pitch period parameter LAG [is] of the previous frame is set to lg [is] (is = 0, ..., NS-1). Here, NS is the number of pitch period parameters included in the previous frame, and when the speech decoder uses the CELP method, the frame is divided into a plurality of subframes, and the pitch period parameter (lag) is coded in each subframe. , The number of subframes is NS. Then, using the past driving sound source of the previous frame, d [n] (n = 0, ..., L-1, L: accumulated driving sound source length), for each lg [is], the phase relation Calculate the number r [is] as follows:

[0020]

[Equation 1] Then, the pitch period lg [is _max ] for is (is _max ) that maximizes r [is] is set as the optimum pitch period of the interpolation frame, and the value of the pitch period parameter LAG [is] of the previous frame coding parameter holder 102 is set. Replace all with LAG [is _max ]. When LAG [is] is coded at a fractional delay interval, the correlation coefficient r is calculated by interpolating multiple samples before and after the delay value lg is to interpolate LAG [is].
It is sufficient to calculate r [is].

Next, the switch 105 is set to the side b, and the preceding frame coding parameter in which the pitch period parameter is replaced with the optimum one is output, and the speech decoder 1 is used.
According to 06, voice decoding is performed on the interpolated frame. The previous frame coding parameter holder 102 and the driving excitation holder 103 are updated on a frame-by-frame basis with the coding parameter and driving excitation used for decoding in the current frame for processing the next frame.

As described above, according to the present embodiment, by selecting the optimum pitch period parameter used during frame interpolation from the plurality of pitch period parameters of the previous frame, driving with more appropriate pitch periodicity can be achieved. A sound source can be generated.

(Embodiment 2) FIG. 2 shows the configuration of a speech decoding apparatus according to a second embodiment of the present invention. In FIG. 2, reference numeral 201 denotes a parameter which is transmitted from the encoder and separates the received data representing the coding information for each frame, which is a fixed section of speech, into respective coding parameters and determines whether the frame is an interpolated frame. Separation / interpolation determiner, 202 is a current frame coding parameter determiner that determines a coding parameter used for speech decoding of the frame from frame interpolation information, 203 is an adaptive codebook that stores past driving sound sources, 204 is a noise sound source, etc. Fixed codebook, which holds a fixed source vector of
Reference numeral 205 denotes a gain decoder that decodes a gain code into a gain value among coding parameters, and 206 denotes an adaptive codebook 203 and a fixed codebook 204 at the time of frame interpolation.
A sound source gain controller that controls the gains of the signals using the signals generated from, 211 is a switch that switches the gain input, and 207 is a multiplier that multiplies the signals generated from the adaptive codebook 203 and the fixed codebook 204 by the gain. , 208 is an adder for generating a driving sound source, 209 is L
An LPC coefficient decoder that decodes PC coefficients, and 210 is a synthesis filter that synthesizes decoded speech by LPC synthesis using a driving sound source as an input.

The operation of the speech decoding apparatus configured as above will be described below. First, the parameter separation / interpolation determiner 201 receives the transmission data transmitted from the encoder, separates it into the coding parameters of the frame, determines whether the frame is an interpolation frame, and determines the coding parameter and Output frame interpolation information. Hereinafter, the case where the frame is an interpolation frame and the case where the frame is not are described separately.

First, when the frame is not an interpolated frame, the current frame coding parameter determiner 202 outputs the output parameter of the parameter separation / interpolation determiner 201 as it is, and the adaptive codebook 203 uses the pitch period parameter (LAG). The adaptive codebook sound source is generated, and the fixed codebook sound source is generated by using the index (IDX) of the fixed codebook 204.
The gain code (GAIN) to the gain decoder 205.
The gain (β, γ) obtained by decoding by
7. The adaptive codebook sound source and the fixed codebook sound source are multiplied by 7 respectively, added by the adder 208 to generate a driving sound source, and the LPC coefficient obtained by the LPC coefficient decoder 209 is used to perform speech decoding by the synthesis filter 210. I do.

Next, the case of the interpolation frame will be described. First, in the current frame coding parameter determiner 202, the coding parameter used for decoding in the previous frame held here is used, the code of the previous frame is used for the LPC code, and the code of the pitch period code is used for the pitch period code.
For the fixed codebook index, the parameter separator 2 is used as the arbitrary pitch period code of the previous frame or the optimum pitch period code obtained by the method described in the first embodiment.
The index code of the current frame obtained from 01 or the index code having a randomly generated value is output. From the adaptive codebook 203 and the fixed codebook 204, the adaptive codebook sound source a [n] (n =
0, ..., N-1) and fixed codebook sound source c [n] (n = 0,
1, ..., N−1) are generated, and using them, the adaptive codebook excitation gain β and the fixed codebook excitation gain γ are obtained by the excitation gain controller 206. β = Ga · DMPa γ = (1−Ga) · (Pow_a / Pow_s) · DMPs where Ga is the power ratio between the adaptive codebook sound source and the fixed codebook sound source that constitute the generated driving sound source. Is set to a constant value (0 <= Ga <= 1), for example, 0.95. DMPa and DMPs are constants (<1) that control power attenuation during continuous interpolation, and Pow_a and Pow
_S is the power of the adaptive codebook sound source a [n] and the fixed codebook sound source c [n], and is represented by the following equation.

[0027]

[Equation 2]

The gains β and γ obtained are used for the switching unit 21.
With 1 being the b side, the adaptive codebook sound source and the fixed codebook sound source are multiplied by the multiplier 207, each sound source is added to generate a driving sound source, and speech synthesis is performed by the synthesis filter 210 to obtain a decoded speech.

As described above, according to this embodiment, the correlation between the adaptive codebook sound source generated from the past driving sound source and having the correlation with the past driving sound source, and the past driving sound source such as the noise component. By controlling the gain so that the power ratio with a fixed codebook sound source having no noise becomes a fixed ratio and generating a driving sound source, it is possible to perform speech decoding in an interpolated frame with suppressed perceptual deterioration.

Further, instead of setting the power ratio Ga between the adaptive codebook sound source and the fixed codebook sound source to a fixed value, a variable value is set according to the correlation of the past driving sound source, and the gain is controlled at a variable ratio. You can In this case, the correlation coefficient of the past driving sound source is obtained, and Ga is increased when the correlation is high and is decreased when the correlation is low. By generating the driving sound source in this way, it is possible to perform voice decoding in an interpolated frame in which deterioration in hearing is further suppressed.

(Embodiment 3) FIG. 3 shows the configuration of a speech decoding apparatus according to a third embodiment of the present invention. The speech decoding apparatus according to the present embodiment has a configuration having coding modes corresponding to the voiced section and the unvoiced section of the speech. In FIG. 3, reference numeral 301 denotes a coder, which separates received data representing coding information for each frame, which is a fixed section of speech, into respective speech coding parameters and decoding mode information, and the frame is an interpolated frame. Parameter separation / interpolation determiner for determining whether or not, 302 is a previous frame encoding parameter holder that retains the encoding parameter / encoding mode used for decoding in the previous frame, and 303 is the speech decoded up to the previous frame. A driving sound source holder that holds a driving sound source signal in a certain section, 304 represents a driving sound source output from the driving sound source holder 303 and a plurality of pitch periods of the previous frame or a parameter similar thereto when the previous frame is in the voiced coding mode. By using the coding mode and voiced coding mode during frame interpolation of the frame. In the case of, the interpolating pitch period determiner for determining the optimum pitch period or a parameter similar thereto, 305 is a mode information switcher for switching the input of the coding mode information at the time of decoding the frame according to the frame interpolation information, and 306 is an interpolation frame A coding parameter switcher that switches the coding parameter used in the frame depending on whether or not there is a mode switcher 307 that switches the coding mode according to the mode information obtained from the output of the mode information switcher, and 308 and 309 are the coding parameters. A voiced mode decoder and an unvoiced mode decoder that perform voice decoding of the frame using the coding parameter and the frame interpolation information selected by the switch 306, 3
Reference numeral 10 is a decoded voice output switch, and 311 is a driving sound source output switch.

The operation of the speech decoding apparatus configured as above will be described below. First, in the parameter separation / interpolation determiner 301, the transmission data transmitted from the encoder is received and separated into each encoding parameter / encoding mode information of the frame, and it is determined whether the frame is an interpolation frame. Then, the coding parameter / coding mode and the frame interpolation information are output. Hereinafter, the case where the frame is an interpolation frame and the case where the frame is not are described separately.

First, when the frame is not an interpolated frame, the coding parameter switch 306 is set to the a side, the mode information switch 305 is set to the a side, and the parameter separator 301 is set.
Using the coding parameter / coding mode of the frame obtained in step 3, the audio decoding is performed according to the coding mode information. That is, voice decoding is performed by the voiced mode decoder 308 when the encoding mode is the voiced mode and by the voiceless mode decoder 309 when the voiced mode is the unvoiced mode. Here, the voiced mode decoder 308 is a mode provided for a voiced section of speech and is a mode in which at least pitch period information (for example, lag information in CELP coding) is used for decoding. Also, the silent mode decoder 309
Is a mode provided for the unvoiced section of the voice, in which the pitch period information is not used for decoding. The decoded audio output switch 310 and the driving sound source output switch 311 are switched according to the decoding mode, and the decoded sound and driving sound source are output according to the mode decoded in the current frame.

Next, the case of the interpolation frame will be described. In the case of an interpolated frame, the mode information switch 305 and the coding parameter switch 306 are respectively set to the b side,
Using the previous frame coding parameter holder 302, the driving sound source holder 303, and the interpolating pitch period determiner 304, the coding mode and coding parameter of the current frame are determined by the following method, and the coding mode is set. The mode switch 307 is switched accordingly, and the voiced mode decoder 308 or the unvoiced mode decoder 309 performs voice decoding using the determined coding parameter.

Here, a method of determining the coding mode / coding parameter in the interpolation frame will be described. The coding mode of the current frame is based on the coding mode of the previous frame. First, when the previous frame coding mode is the voiced mode, it is determined that the current frame is the voiced mode, and a plurality of pitch period parameters of the previous frame held in the previous frame coding parameter holder 302 (in the case of CELP type decoder, , The lag of each sub-frame of the previous frame is input, and the interpolation pitch cycle determiner 304 determines the optimum interpolation pitch cycle that maximizes the autocorrelation of the driving sound source by the method described in the first embodiment. Pitch period parameter of. Instead of obtaining the optimum pitch period parameter for the current frame, any parameter of the previous frame may be used. Even if the previous frame is in the voiced mode, if the normalized autocorrelation coefficient in the determined optimum pitch period is less than a certain threshold value, the pitch periodicity is considered to be low, and the unvoiced mode is determined. May be. Next, when the previous frame coding mode is the unvoiced mode, the voiced / unvoiced mode is determined based on the correlation between the past driving sound sources. That is, using the past driving sound source held by the driving sound source holder 303, the interpolation pitch cycle determiner 30 is used.
In 4, the delay value that maximizes the autocorrelation coefficient of the driving sound source, that is, the pitch period is obtained, and if the normalized correlation coefficient at that time is greater than or equal to a threshold value, the pitch period is output as the optimum value. , The interpolated frame is determined to be the voiced mode, and conversely, when it is less than or equal to the threshold value, it is determined to be the unvoiced mode.

As described above, according to this embodiment, the coding mode of the current frame is determined, the mode information is output, and in the case of the voiced mode, the preceding frame coding is performed including the optimum pitch period. The parameter is output as the decoding parameter of the current frame.

Instead of the interpolated frame coding mode determination as described above, the interpolated pitch cycle determiner 304 determines the past decoded driving sound source regardless of whether the previous frame is the voiced coding mode or the unvoiced coding mode. Then, the pitch period that maximizes the autocorrelation is obtained, and when the normalized correlation coefficient at that time is greater than or equal to a threshold value, the pitch period is output as the optimum value, and the interpolated frame is determined to be the voiced mode, On the contrary, when it is equal to or less than the threshold value, it may be determined to be the unvoiced mode.

[0038]

As is apparent from the first embodiment, the present invention is more suitable by selecting the optimum pitch period parameter used during frame interpolation from the plurality of pitch period parameters of the previous frame. It is possible to generate a driving sound source having pitch periodicity.

As is apparent from the second embodiment, the correlation between the adaptive codebook sound source generated from the past driving sound source and having the correlation with the past driving sound source, and the past driving sound source such as the noise component. The power ratio with a fixed codebook sound source that has no effect is controlled by controlling the gain at a fixed ratio or at a variable ratio according to the correlation of the past driving sound source to generate a driving sound source. It is possible to perform voice decoding with suppressed interpolation frames.

Further, as is clear from the third embodiment, the interpolated pitch period determiner evaluates the voicing / unvoicedness of the previous frame based on the correlation of the past driving sound sources, and when the voicing is high. In the voiced mode, otherwise, by performing the decoding process of the interpolation frame in the unvoiced mode, it is possible to perform the voice decoding in which the deterioration in the auditory sense is further suppressed.

[Brief description of drawings]

FIG. 1 is a block diagram of a speech decoding apparatus according to a first embodiment of the present invention.

FIG. 2 is a block diagram of a speech decoding apparatus according to a second embodiment of the present invention.

FIG. 3 is a block diagram of a speech decoding apparatus according to a third embodiment of the present invention.

FIG. 4 is a block diagram of a speech decoding device in a conventional example.

[Explanation of symbols]

101 Parameter separation / interpolation judgment device 102 previous frame coding parameter holder 103 Drive sound source holder 104 Interpolation pitch cycle determiner 105 switch 106 speech decoder 201 Parameter separation / interpolation judgment device 202 previous frame coding parameter holder 203 Adaptive codebook 204 fixed codebook 205 gain decoder 206 gain controller 207 multiplier 208 adder 209 LPC coefficient decoder 210 Synthesis filter 211 Switch 301 Parameter separation / interpolation judgment device 302 previous frame coding parameter holder 303 Drive sound source holder 304 Interpolation pitch cycle determiner 305 Mode information switch 306 Encoding parameter switch 307 Mode switch 308 Voiced Mode Decoder 309 Silent mode decoder 310 Decoded audio output switch 311 Drive sound source output switch

   ─────────────────────────────────────────────────── ─── Continued front page       (56) References JP-A-3-51900 (JP, A)                 JP-A-5-73097 (JP, A)                 Japanese Patent Laid-Open No. 5-19796 (JP, A)

Claims (12)

(57) [Claims] 1. Separation of received data, which is transmitted from an encoder and represents coding information for each frame, which is a constant section of speech, into respective coding parameters and also determines whether or not the frame is an interpolated frame.・
An interpolation determiner, a previous frame coding parameter holder that holds the coding parameter used for decoding in the previous frame,
A driving sound source holder that holds a driving sound source signal in a certain section of the sound decoded up to the previous frame, and a plurality of pitches of the driving sound source signal from the driving sound source holder and the preceding frame from the preceding frame coding parameter holder Using a cycle parameter, from among the plurality of pitch cycle parameters of the previous frame, an interpolation pitch cycle determiner for determining an optimum pitch cycle parameter for frame interpolation of the frame,
A switching device that switches the coding parameter used in the frame depending on whether it is an interpolated frame or not, and a speech decoder that performs speech decoding of the frame using the coding parameter and frame interpolation information selected by the switching device are provided. Speech decoding device. 2. The speech decoder according to claim 1, wherein the speech decoder is a CELP type decoder, and a lag for each subframe of the previous frame is used as the plurality of pitch period parameters input to the interpolating pitch period determiner. Device. 3. The interpolating pitch cycle determiner outputs a pitch cycle that maximizes the autocorrelation among a plurality of past pitch cycle parameters for the past decoded driving sound source as an optimum value. Voice decoding device. 4. Parameter separation for separating received data, which is transmitted from an encoder and represents coding information for each frame, which is a fixed section of speech, into respective coding parameters and determines whether or not the frame is an interpolation frame.・
Interpolation determiner, current frame coding parameter determiner that determines coding parameters used for speech decoding of the frame from frame interpolation information, adaptive codebook that stores past driving sound sources, fixed sound source vector such as noise sound source A fixed codebook that holds, a source gain controller that controls the gains of the adaptive codebook and the signals generated from the fixed codebook during frame interpolation, and a signal generated from the adaptive codebook and the fixed codebook. A speech decoding apparatus including a multiplier that multiplies a gain and an adder that generates a driving sound source. 5. The speech decoding apparatus according to claim 4, wherein the excitation gain controller controls the gain so that a ratio of powers of the adaptive codebook excitation and the fixed codebook excitation becomes a fixed ratio. 6. A sound source gain controller varies the power ratio of each of the adaptive codebook sound source and the fixed codebook sound source according to the correlation of past driving sound sources, and when the correlation is high, the adaptive codebook sound source is adjusted. 5. The speech decoding apparatus according to claim 4, wherein the gain is controlled so that the power ratio of is increased. 7. A speech decoding apparatus having coding modes corresponding to voiced sections and unvoiced sections of speech, respectively.
Parameter that separates the received data that is transmitted from the encoder and that represents the coding information for each frame, which is a fixed section of speech, into each speech coding parameter and coding mode information, and that determines whether the frame is an interpolated frame Separation / interpolation decision unit, previous frame coding parameter holder that holds the coding parameter / coding mode used for decoding in the previous frame, and driving excitation signal in a certain section of the sound decoded up to the previous frame A driving sound source holder for driving, and a driving sound source signal output from the driving sound source holder and a plurality of pitch period parameters of the previous frame when the previous frame is in a voiced coding mode, the code at the time of frame interpolation of the frame. If the encoding mode is determined and the encoding mode is the voiced encoding mode, the previous frame Number of pitch cycle parameters, an interpolation pitch cycle determiner that determines an optimum pitch cycle parameter, a mode information switch that switches input of coding mode information at the time of decoding the frame according to frame interpolation information, and an interpolation frame A coding parameter switcher that switches a coding parameter used in the frame depending on whether or not the mode, a mode switcher that switches a coding mode based on the mode information obtained from the output of the mode information switcher, and the coding parameter switcher. A voice decoding apparatus including a voiced mode decoder and a voiceless mode decoder for performing voice decoding of the frame using the coding parameter and frame interpolation information selected in. 8. The voiced mode decoder and the unvoiced mode decoder are both CELP type decoders, the voiced mode decoder has at least an adaptive codebook, and pitch period information is used for decoding, and the unvoiced mode decoder is used. 8. The speech decoding apparatus according to claim 7, further comprising: a configuration having no adaptive codebook, wherein lags for each subframe of the previous frame are used as the plurality of pitch period parameters input to the interpolation pitch period determiner. 9. The interpolation pitch period determiner determines a pitch period that maximizes the autocorrelation among a plurality of past pitch period parameters for a past driving sound source when the previous frame is in a voiced coding mode. The speech decoding according to claim 7, wherein the decoded frame is output as an optimum value and the interpolated frame is determined as a voiced mode.
Apparatus. 10. A pitch period which maximizes an autocorrelation among a plurality of past pitch period parameters for a past decoding driving excitation when the previous frame is in a voiced coding mode. If the normalized correlation coefficient at that time is more than a certain threshold value, the pitch period is output as an optimum value and the interpolated frame is determined to be a voiced mode. The speech decoding apparatus according to claim 7, wherein the determination is performed. 11. The interpolation pitch period determiner obtains a pitch period that maximizes the autocorrelation with respect to the past driving sound source when the previous frame is in the unvoiced coding mode, and the normalized correlation coefficient at that time is determined. 8. The speech decoding apparatus according to claim 7, wherein when the threshold value is equal to or more than a threshold value, the pitch period is output as an optimum value to determine the interpolated frame as a voiced mode, and conversely, when the threshold value is less than or equal to the threshold value, the unvoiced mode is determined. . 12. An interpolating pitch period determiner obtains a pitch period that maximizes autocorrelation with respect to a past driving sound source regardless of a voiced coding mode / unvoiced coding mode of a previous frame, 8. If the normalized correlation coefficient is above a certain threshold value, the pitch period is output as an optimum value to determine the interpolated frame as a voiced mode, and conversely, if it is below the threshold value, it is determined as an unvoiced mode. The described speech decoding device.