JP3262652B2 - Audio encoding device and 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 coding apparatus and a speech decoding apparatus, for example, a VSELP (Vector Sum E
xcited Linear Prediction: Vector addition excitation linear prediction) This is suitable for application to a speech coding system.

[0002]

2. Description of the Related Art VSELP speech coding is a type of code-excited linear prediction (CELP) speech coding,
The voice signal is parameter-encoded into vocal tract and vocal cord (excitation source) information. In particular, the VSELP speech coding method differs from the CELP coding method in the parameter coding method of the excitation source. This VSELP speech coding method is described in, for example, US TIA (Telecommunications Ind.
ustry Association) committee has standardized it as a voice coding system for digital cellular.

[0003] Such a VSELP speech coding system is described in, for example, References A and B.

Reference A: "VECTOR SUM EXCITED LINEAR PR
EDICTION (VSELP) SPEECH CODINGAT 8KBPS ”, Ira AG
erson and Mark A. Jasiuk, Proc. IEEE Inc. Conf. on A
coustics, Speech and Signal Processing, pp.461-46
4, April 1990 Document B: Japanese Patent Laid-Open No. 4-190399 FIG. 2 shows a functional configuration of a conventional speech coding apparatus according to the VSELP speech coding method, mainly in accordance with the contents described in Document A.

[0005] In FIG. 2, the information related to the dashed output lines drawn from each functional block indicates information transmitted to the audio decoding device, and the information related to the solid output lines drawn from each functional block. The information is information processed when determining (searching) information to be transmitted.

The VSELP speech coding apparatus includes an input speech analysis unit (which can be further divided into a speech power calculation unit, a speech power interpolation unit, and a vocal tract analysis unit) 101 and a target signal creation unit 102. , A long-term lag selector 103, a first code selector 104, a second code selector 105, a gain selector 106, three amplifier circuits 107 to 109, and an adder circuit 110. The apparatus receives a digitized audio signal V sampled at a predetermined frequency and performs encoding.

[0007] The input voice analysis unit 101 receives the input voice signal V
Is subjected to LPC analysis (a type of vocal tract analysis method) for every 160 samples (one frame),
A coefficient k and a voice average power R are obtained, and the LPC coefficient k and the voice average power R are quantized and interpolated with respect to a subframe including 40 samples.
LPC coefficient kj and average sound power R for each subframe
Outputs 0.

The method of interpolating the average voice power for such a subframe is as follows. The average voice power of the first subframe is the average voice power of the previous frame, and the average voice power of the second subframe is the average voice power of the previous frame. The average of the power and the average power of the voice of the current frame
And the average voice power of the fourth subframe is the average voice power of the current frame.

The target signal generator 102 generates a residual signal of the input voice signal V by an analysis filter using the LPC coefficient kj for each subframe,
By inputting the residual signal to a weighting filter using the PC coefficient kj to resynthesize the speech signal, and removing the influence of the excitation signal (excitation code vector) ex synthesized in the previous subframe, Generate and output a target signal p.

[0010] With the target signal p as a target, each function unit described later searches and determines excitation source information related to its own function unit. It is assumed that the codes indicating various types of information shown in FIG. 2 are in a determined state, and the search (candidate) being searched for will be described with the code t appended to the end of the code.

[0011] The long-term lag selecting section 103 updates and saves the excitation signal ex of a plurality of subframes that have been synthesized before each time processing on a subframe basis is completed (internal value long-term filter state). ). The long-term lag selecting unit 103 extracts an excitation signal c0t of 40 samples each with various lags Lt from the stored combined excitation signals (the excitation signal of each 40 samples is a long-term excitation code vector. ),
A lag-corresponding long-term excitation code vector c0t is added to the weighting filter using the LPC coefficient kj.
Is calculated, and the square error of the difference between the signal f0t obtained by inputting the target signal and the target signal p is calculated, and a long-term excitation code vector c0 that minimizes the square error is obtained and output. Note that a lag Lt related to the long-term excitation code vector corresponds to each long-term excitation code vector c0t as an index.

The first code selection unit 104 stores a plurality of fixed data as a basis vector (basis vector) v1. The first code selection unit 104 creates a plurality of first excitation code vectors u1t by combining a parameter (gray code) θ that gives the vector a positive / negative polarity with respect to the basis vector (refer to Expression (1) in Document A).
After applying a weighting filter using the LPC coefficient kj to each excitation code vector u1t, a signal f1t orthogonalized to the long-term excitation code vector c0 selected by the long-term lag selection unit 103 is further created. And a target signal p, and calculates and outputs a first excitation code vector c1 (u1) that minimizes the square error. Note that an index It is assigned to each first excitation code vector u1t.

The second code selector 105 stores a plurality of basis vectors v2 different from those of the first code selector 104. The second code selecting unit 105 creates a plurality of second excitation code vectors u2t by combining the parameter θ with the basis vector, applies each excitation code vector to a weighting filter using the LPC coefficient kj, and then performs long-term excitation. A signal f2t that is orthogonalized to the code vector c0 and that is orthogonalized to the optimal first excitation code vector c1 selected by the first code selection unit 104 is created. Is calculated, and the second excitation code vector c2 (u2) that minimizes the square error is obtained and output. The index Jt is assigned to each second excitation code vector u2t.

The gain selection unit 106 includes amplification circuits 107 to
109 and the addition circuit 110, the optimum gains β, γ for the respective excitation code vectors c0, c1, c2.
1, γ2.

The gain selector 106 includes gains β, γ1, γ2 for the respective excitation code vectors c0, c1, c2.
Are stored as a set of parameters GS, P0, and P1 obtained by equivalently converting the gains β, γ1, and γ2. A set of parameters GS, P0, and P1 is vector-quantized, and an index (gain index) G is given to each set. Note that the gains β and γ
The relationship between the set of 1, γ2 and the set of parameters GS, P0, P1 is described in the above-mentioned document A in equations (15) to (23).

The gain selector 106 has a parameter GS
For each set of t, P0t, and P1t, the LPC coefficient kj,
Average voice power R0 and each excitation code vector c0, c
1, c2, and converted to gains βt, γ1t, and γ2t and provided to the corresponding amplifier circuits 107, 108, and 109. By the amplification circuits 107 to 109 and the addition circuit 110, a combined excitation signal ext which is a product-sum operation result of each of the excitation code vectors c0, c1, c2 and the corresponding gain βt, γ1t, γ2t is obtained as a plurality of candidates. The gain selection unit 106 calculates LP for each synthesized excitation signal ext.
A local reproduction signal is obtained by applying a synthesis filter obtained from the C coefficient kj, a square error of a difference between each local reproduction signal and the target signal p is calculated, and gains β, γ1, γ2 for minimizing the square error are calculated. Are determined to be optimal.

Optimum excitation code vectors c0, c1, c
2 and the combined excitation signal ex generated from the optimal gains β, γ1, and γ2 are used for the synthesis of the target signal for the next subframe, the update of the long-term filter state, and the like.

The VSELP speech coding apparatus comprises an LPC coefficient kj, an average speech power R0, an index (long term lag) L of a long-term excitation code vector c0, an index I of an optimal first excitation code vector c1, an optimal second The index J of the excitation code vector c2 and the index G of the set of the optimal gain parameters GS, P0, P1 are transmitted to the VSELP speech decoder.

Next, although the illustration of the VSELP speech decoder is omitted, its decoding operation will be briefly described.

In the VSELP speech decoding apparatus, the received lag L is applied as an index to the long-term filter state updated based on the synthesized excitation signals of several past subframes obtained as described later. To decode the long-term excitation code vector c0. Also, the same stored basis vector as the encoding device (if the encoding device and the decoding device are mounted on the same device, those prepared by the encoding device) v
1 and the index I of the received first excitation code vector, and decodes the optimal first excitation code vector c1, and stores the same stored basis vector (the same as the encoding device and the decoding device) An optimal second excitation code vector c2 is decoded from v2 and the index J of the received second excitation code vector if it is installed in the apparatus. Further, a set of optimal gain parameters GS, P0, P1 is extracted from the optimal gain index G, and the obtained long-term excitation code vector c0, first excitation code vector c1, second excitation code vector c2, and received LPC The optimum gain β, γ1, γ2 is decoded from the coefficient kj and the average audio power R0.

Then, gains β, γ1, γ2 corresponding to the obtained long-term excitation code vector c0, first excitation code vector c1, and second excitation code vector c2 are obtained.
, And add them to obtain a combined excitation signal ex.
By passing through the synthesis filter unit configured using
The audio signal V is reproduced, and the noise component of the audio signal V is compressed and output through a post-filter unit configured using the received LPC coefficient kj.

By the way, the VSELP speech coding method is as follows.
US TIA (Telecommunications Industry Association)
n) It is mainly used for compression communication of audio signals such as digital mobile communication, as standardized by the committee as an audio coding method for digital cellular.

For this reason, a processing unit (speech encoding device or speech decoding device) that executes the VSELP speech encoding method
Requires the smallest possible size and low power consumption. In order to meet such a demand, it is effective to fix the bit length (word length) of data handled in the processing device as short as possible and to execute the arithmetic processing performed by the processing device in a fixed-point representation. It is. Therefore, the conventional VSELP speech coding apparatus shown in FIG.
In the SELP speech decoding apparatus, the arithmetic processing is executed in a fixed-point representation.

However, if the bit length of the data to be processed is selected to be short and the arithmetic processing in the fixed-point representation is employed, if the average audio power of the input audio signal of the subframe at that time changes, the arithmetic operation is performed. The values of various internal variables used for processing also change, and in the selected fixed-point representation, some variables may cause overflow, etc., resulting in a decrease in calculation accuracy and a significant decrease in the quality of the final decoded speech signal. Had also been caused.

The above-mentioned Document B describes an invention which can solve the inconvenience caused by calculating short bit length data in fixed point representation. That is, attention is paid to the fact that there is a specific relationship between the magnitude (average power) of the input audio signal and the magnitude of the internal variable used for the arithmetic processing, and the arithmetic processing is performed in accordance with the determined average audio power R0. It describes that a decimal point position of an internal variable or the like to be used is switched, and a predetermined operation is performed at a fixed point position based on the switched decimal point position.

In FIG. 2, the target signal generator 10
2. Long term lag selection unit 103, first code selection unit 104, second code selection unit 105, and gain selection unit 10
Reference numeral 6 denotes a functional unit that executes various operations, and is a unit to which the method described in Document B is applied. In FIG. 2, a target signal creating unit 102, a long term lag selecting unit 10
3. First code selector 104 and second code selector 10
The fact that the audio average power R0 is input in parentheses for the fifth functional block indicates that the method of Document B is applicable. Also in the gain selection unit 114, not only the case where the gain parameter is converted into the gain, but also the average voice power R0 can be used to determine the fixed-point position.

When the sound average power R0 is input to each of these functional units, the stage to which the sound average power R0 belongs is detected, and for example, a table prepared in advance is referred to and the variable used for the calculation is determined. Get fixed-point representation. After such a process, a predetermined operation (for example, a process of passing a predetermined signal through a filter) is performed.

FIG. 3 shows an example of the structure of a table for switching the fixed-point position. For example, if the audio average power R0 is equal to or more than the value PW0 and less than PW1, the variable 1
, The fixed-point position of variable 2 is set to a2, and the fixed-point position of variable y is set to ay. Incidentally, a table in which the stage to which the audio average power R0 belongs and the fixed-point representation of the internal variable are associated is, for example, an output signal from the functional unit corresponding to the stage to which the audio average power R0 belongs (also an internal variable). The fixed-point position of an internal variable that can realize such a fixed-point position of the output signal is stored in consideration of the fixed-point position of the output signal.

[0029]

However, among the plurality of functional units in which the fixed-point representation of the internal variables can be switched, the long term lag selecting unit 103 has a different characteristic from other functional units. That is, the long-term lag selecting unit 103 stores the combined excitation signal ex of several past subframes as a long-term filter state, and depending on the stored subframe, the fixed-point position of the combined excitation signal may be different. Different from the others. When the fixed-point position of the long-term excitation code vector c0 from the long-term lag selecting unit 103 is determined based on the audio average power R0 of the current target subframe, a long-term filter is added to the determined fixed-point position. It is necessary to match the fixed-point position of the combined excitation signal stored as a state and extract it. This means that data is shifted, and the value may overflow depending on the shift direction.

That is, even if the fixed-point position is switched based on the average voice power R0, the long term lag selecting section 103 cannot obtain sufficient accuracy.

As described above, not only gain selection but also
For example, as described above, the average audio power R0 of each subframe used for switching the fixed-point position is, as described above, the average audio power of the first subframe is the average audio power of the previous frame, and the average audio power of the second subframe is The average audio power of the previous frame and the average audio power of the current frame are determined as the geometric mean, and the average audio power of the third and fourth subframes is determined to be the average audio power of the current frame. It should be noted that the average information of the two average audio powers is obtained by geometric mean for the second subframe because the average audio power is the sum of squares of the sample values, and This is because it is closer to the sum of squares of the sample sequence of the intermediate value than the averaging.

According to the above-mentioned interpolation method, the average voice power of the first, third and fourth sub-frames can be easily obtained, but the second sub-frame requires calculation. In addition, since it is a geometric mean operation, multiplication and squaring are required. In general, the process of finding the square root depends on a complicated calculation mechanism and requires many processing steps. Therefore, this is contrary to the miniaturization and low power consumption as much as possible which are required for a device (a voice coding device or a voice decoding device) to which the VSELP voice coding method is applied.

Although the problem has been described above with respect to the device that complies with the VSELP speech coding method, a device that complies with the general CELP speech coding method also outputs a certain kind of excitation code vector storing the past synthesized excitation signal. adaptation excited
Vibration code vector selection unit (long-term lag election described above
(Corresponding to the selection unit), and there is a type that interpolates the average voice power for each subframe. That is,
The above problem also occurs in a device that follows a general CELP speech coding scheme.

The present invention has been made in view of the above points, and realizes an improvement in accuracy by switching the fixed-point position even in an adaptive excitation code vector selection unit storing past synthesized excitation signals. It is an object of the present invention to provide a speech encoding device and a speech decoding device capable of performing the above-mentioned operations.

Further, the present invention provides a speech coding apparatus capable of executing, by a simple configuration or a simple process, obtaining an average speech power per subframe from a speech average power per frame. Is what I tried.

[0036]

In order to solve this problem, according to the first aspect of the present invention, a voice signal is parameter-encoded by digital processing using a finite word length.
An audio encoding device according to an ELP audio encoding method,
A fixed internal variable that stores a combined excitation signal of several past frames (a concept including a subframe in this claim) and outputs a part of the signal sequence as an adaptive excitation code vector component. The following configuration is provided in a speech coding apparatus having an adaptive excitation code vector selection unit capable of switching a decimal point position.

That is, for each of the combined excitation signals of the past several frames stored in the adaptive excitation code vector selection unit, a decimal point position buffer unit for storing fixed-point position information, and stored in the decimal point position buffer unit A fixed-point position selector for determining a fixed-point position of an internal variable in the adaptive excitation code vector selector based on the plurality of fixed-point position information and the average voice power for the current frame.

According to the second aspect of the present invention, an audio signal is
An audio encoding apparatus according to a CELP audio encoding method for performing parameter encoding by digital processing with a finite word length, comprising: an audio power calculation unit for obtaining an average audio power of each frame of an audio signal; A speech encoding apparatus having a speech power interpolator for obtaining speech average power information for each subframe obtained by equally dividing one frame into several frames from the speech average power is as follows.

That is, the audio power interpolator has at least a means for quantizing the average audio power and, when obtaining the average audio power information for each sub-frame, treats the average audio power for each frame by a quantized index and For the subframes for which the geometric average information of the audio average power of the frames is obtained, the arithmetic average of the indexes is obtained, the index obtained for each subframe is output, or the obtained index is inversely quantized. And output the average power of the voice returned via the.

Furthermore, the present invention according to claim 3 is a speech encoding apparatus according to a CELP speech encoding method for encoding a speech signal by digital processing with a finite word length, wherein the speech signal is averaged for each frame of the speech signal. A voice power calculator for obtaining power; a voice power interpolator for obtaining voice average power information for each sub-frame obtained by equally dividing one frame into several frames from the voice average power of two consecutive frames; An adaptive excitation code vector selector capable of switching a fixed-point position of an internal variable, storing a composite excitation signal of the sub-frames and outputting a part of the signal sequence as an adaptive excitation code vector component The speech encoding device was as follows.

That is, the audio power interpolating unit includes at least audio average power quantizing means, and when obtaining audio average power information for each sub-frame, treats the audio average power for each frame with a quantized index, and For the subframes for which the geometric average information of the audio average power of the frames is obtained, the arithmetic average of the indexes is obtained, the index obtained for each subframe is output, or the obtained index is inversely quantized. And a fixed-point position buffer unit for storing fixed-point position information for each of the past several sub-frame synthesized excitation signals stored in the adaptive excitation code vector selection unit, And a plurality of fixed point position information stored in the decimal point position buffer section and the current subframe. And a voice average power of beam, provided with decimal point selection unit which determines a fixed-point position of the internal variables of the adaptive excitation code vector selection unit.

According to a fourth aspect of the present invention, there is provided a speech encoding apparatus according to any one of the first to third aspects of the present invention.
The ELP speech encoding method is a forward-type VSELP speech encoding method, which is one type of the ELP speech encoding method.

According to a fifth aspect of the present invention, there is provided a speech decoding apparatus according to the CELP speech coding system corresponding to the speech coding apparatus according to the first aspect of the present invention. Is a concept that includes subframes), and outputs a part of the signal sequence as an adaptive excitation code vector component. Selects an adaptive excitation code vector that can switch the fixed-point position of internal variables. The following configuration is provided in the audio decoding device having the section.

That is, for each of the combined excitation signals of the past several frames stored in the adaptive excitation code vector selection unit, a decimal point position buffer unit for storing fixed-point position information, and a decimal point position buffer unit A fixed-point position selector for determining a fixed-point position of an internal variable in the adaptive excitation code vector selector based on the plurality of fixed-point position information and the average voice power for the current frame.

According to a sixth aspect of the present invention, there is provided a speech decoding apparatus according to any one of the fifth aspects of the present invention.
P-speech coding method is one of the forward type VS
It is characterized by the ELP audio coding method.

[0046]

According to the first and fifth aspects of the present invention, there are provided a speech encoding apparatus and a speech decoding apparatus which conform to a CELP encoding system for encoding a speech signal by digital processing with a finite word length and transmitting the encoded signal. It has an adaptive excitation code vector selector capable of switching a fixed-point position of an internal variable, storing a composite excitation signal of several frames and outputting a part of the signal sequence as an adaptive excitation code vector component. About things.

There has already been proposed a method of switching the fixed-point position of an internal variable used when each functional unit performs processing according to the average voice power. In the adaptive excitation code vector selection unit that saves and uses the combined excitation signals of the past several frames, the fixed-point position differs for each of the combined excitation signals of the past several frames. If the fixed-point position is determined from the average voice power of the fixed frame, the fixed-point position may not be appropriate in view of the synthesized excitation signal of the past frame (for example, even if a small number can be accurately represented, a large number overflows. May be incorrect).

Therefore, according to the first and fifth aspects of the present invention, a decimal point position buffer section and a decimal point position selecting section are provided, and the decimal point position selecting section includes a plurality of fixed decimal point positions stored in the decimal point position buffer section. The fixed-point position of the internal variable in the adaptive excitation code vector selector is determined from the information and the average voice power of the current frame. Of course, in the speech encoding device and the speech decoding device, the adaptive excitation code vector selection unit must output the adaptive excitation code vector in the same state, and the speech encoding device requires the decimal point position buffer unit and the decimal point position selection unit. When such units are provided, these functional units are also provided in the corresponding audio decoding device.

According to a second aspect of the present invention, a voice signal is parameter-encoded by digital processing using a finite word length.
An audio encoding device according to an ELP audio encoding method,
An audio power calculation unit for calculating an average audio power of each frame of an audio signal, and audio for obtaining average audio power of each sub-frame obtained by equally dividing one frame into several pieces from the average audio power of two consecutive frames. It is assumed that the speech encoding device has a power interpolation unit.

For some subframes, average information of the average sound power of two consecutive frames is obtained as interpolation information. In consideration of the fact that the average voice power is obtained by the sum of squares of the sample values, interpolation has conventionally been performed using the geometric mean value. However, the interpolation configuration or interpolation processing is complicated.

Therefore, in the present invention of claim 2, the audio power interpolating unit is provided with at least audio average power quantizing means, and when obtaining the audio average power information for each sub-frame, the audio power for each frame is obtained. The average power is treated as a quantized index. In some subframes, the geometric average information of the audio average power of the two frames may be obtained. In this case, the arithmetic average of the indices is used. Depending on how to use the audio average power information for each sub-frame, an index of the audio average power obtained for each sub-frame may be output, and the obtained index may be output via the inverse quantization means. The returned audio average power may be output.

According to a third aspect of the present invention, there is provided a speech encoding apparatus having both the features of the first and second aspects of the present invention.

According to a fourth aspect of the present invention, there is provided a speech encoding apparatus according to any one of the first to third aspects of the present invention.
The ELP audio coding system is limited to one of the forward-type VSELP audio coding systems, and the present invention according to claim 6 is the audio decoding apparatus according to claim 5 of the present invention. Is limited to the forward VSELP speech coding method, which is one type of the CELP speech coding method. In compressed communication such as mobile communication, a forward type VSELP speech coding system is adopted, and any of the present inventions of claims 1 to 3 and the present invention of claim 6 are also effective inventions. .

[0054]

【Example】

(A) VSELP Speech Encoding Apparatus Hereinafter, an embodiment of a speech encoding apparatus according to the present invention will be described in detail with reference to the drawings. This embodiment is an example of a VSELP speech coding apparatus according to the VSELP speech coding method, and FIG. 1 shows a functional block configuration thereof.

In FIG. 1, the information on the dashed output lines drawn from each functional block indicates information transmitted to the decoding device, and the information on the solid output lines drawn from each functional block. The information is information processed when determining (searching) information to be transmitted.

Also, it is assumed that the codes indicating the various information shown in FIG. 1 are in the determined state, and the searching (candidate) will be described with the code t appended to the end of the code.

As shown in FIG. 1, the VSELP speech coding apparatus of this embodiment has a speech power calculator 201, a speech power interpolator 202, an LPC analyzer 203, a target signal generator 204, and a long term lag selector 205. , Decimal point position buffer 206, decimal point position selector 207,
It comprises a first code selector 208, a second code selector 209, a gain selector 210, three amplifier circuits 211 to 213, and an adder circuit 214.

The audio power calculator 201 is supplied with the digital input audio signal V. The audio power calculator 201 converts the average audio power of the input audio signal V into a frame (for example, 16 bits).
The calculation is performed for each sample (for example, 0 samples) (for example, the sum of squares of sample values), and the average audio power Rf for each frame is provided to the audio power interpolation unit 202 and the LPC analysis unit 203.

The audio power interpolation unit 202 interpolates the average audio power R0 of each subframe (for example, 40 samples) from the average audio power Rf of the current frame and the previous frame and obtains the average audio power R0 of each subframe. Output.

Here, the audio power interpolation unit 202 has a quantization table 202a of audio average power expressed in logarithmic order, and obtains an index indicating to which quantization stage the audio average power belongs. It has been made possible. Also, the audio power interpolation unit 202
Also has an inverse quantization table 202b for obtaining the average voice power from the index of the average voice power.

In this embodiment, the audio power interpolation unit 202 converts the average audio power Rf of each frame into an index using the quantization table 202a, and performs an interpolation process on the sub-frame at the index stage.
The index of the average audio power of each subframe is converted to the average audio power R0 using the inverse quantization table 202b.

The interpolation of the index for each sub-frame is as follows:
For example, the following is performed. (1) The index of the average audio power of the previous frame is used as the index of the average audio power of the first subframe. (2) The arithmetic average of the index of the average audio power of the previous frame and the index of the average audio power of the current frame is used as the index of the average audio power of the second subframe. (3) The index of the average voice power of the current frame is used as the index of the average voice power of the third subframe. (4) The index of the average voice power of the current frame is used as the index of the average voice power of the fourth subframe.

Note that the inverse quantization table 202b is configured more finely than the quantization table 202a so that an index having a fraction obtained by arithmetic averaging can be converted into audio average power.

In the interpolation method of this embodiment, the conversion of the average voice power into the index and the conversion from the index into the average voice power are performed using the tables 202a and 202b, so that the processing is simple.
The process for obtaining the average information of the individual voice average powers is also a simple process because the arithmetic mean is used instead of the geometric mean.
The geometric mean is obtained by multiplication and squaring, while the arithmetic mean is obtained by addition and halving. The addition in the arithmetic averaging operation is also a simple process, and is a simple process of shifting half the data of a plurality of bits by one bit.

In the above description, the index is inversely quantized and returned to the audio average power, and then applied to various functional units. However, the index may be applied to various functional units as it is. You may make it transmit to a decoding apparatus. As will be described later, the audio average power R0 for each subframe is used for the fixed-point position of an internal variable in various functional units, and its absolute value is not a problem. It is sufficient if it is clear, and it is sufficient to output the index. Rather, the format in which the index is provided simplifies the configuration of various functional units. However, the gain selection unit 2
In 10, since the absolute value of the average sound power of each sub-frame is required for the gain selection operation, it is necessary to provide an inverse quantization table in the gain selection unit 210 when giving an index. Become. However, in the following, description will be made assuming that the average sound power R0 per subframe is given to various functional units.

The LPC analysis section 203 is provided as vocal tract analysis means. LPC analysis section 203 switches the fixed-point positions of variables in the LPC analysis operation performed internally based on the given average voice power Rf of the current frame (see FIG. 3). Thereafter, an LPC analysis is performed on the input voice signal V for one frame to obtain an LPC coefficient kj.
Is output.

In the prior art, the LPC coefficient is calculated and output for each sub-frame, but in this embodiment, the LPC coefficient kj is determined for each frame.
Is used in all subframes. However, this difference is irrelevant to the feature of the present invention, and the LPC coefficient may be obtained and output for each subframe as in the related art.

The target signal generator 204 performs substantially the same processing as the conventional target signal generator 102 shown in FIG. 2 to form and output a target signal p for each subframe. The difference from the conventional target signal creation unit 102 shown in FIG. 2 is that before the start of the target signal p creation calculation processing, the target signal p creation calculation is performed in accordance with the average audio power R0 of the input subframe. Is to set the fixed-point position of the variable used for.

The long-term lag selecting section 205 performs almost the same processing as the conventional long-term lag selecting section 103 shown in FIG. 2 to obtain and output the optimal long-term excitation code vector c0 for each subframe. Of course, at this time, the optimum long term lug L is also determined.

However, the long-term lag selecting section 205 of this embodiment has a function of switching the fixed-point position of an internal variable used for determining the optimum long-term excitation code vector c0 and the optimum long-term lag L. Is different from the conventional one, and the method of switching the fixed-point position of the internal variable is also different from the method described in Document B based only on the average sound power R0.

For switching the fixed-point position, the long-term lag selecting unit 205 of this embodiment is provided with a decimal-point position buffer unit 206 and a decimal-point position selecting unit 207.

FIG. 4 shows the contents stored in the long-term filter state buffer unit in the long-term lag selection unit 203 and the decimal point position buffer unit 206.

The long-term lag selecting section 203 is supplied with the final combined excitation signal ex for each subframe, and the final combined excitation of several immediately preceding subframes (four in FIG. 4 as an example). Signals ex (-1), ex (-2), ex
(-3) and ex (-4) are as shown in FIG.
It is stored in the long-term filter state buffer. At the time of storing the combined excitation signal ex, information on the fixed-point position fp related to the combined excitation signal is given to the decimal-point position buffer unit 206, and as shown in FIG. Position fp (-
1), fp (−2), fp (−3), and fp (−4) are stored in the decimal point position buffer unit 206.

Here, when searching for the optimal long-term excitation code vector c0 for the current subframe,
It is necessary to extract data for one subframe period while changing the long term lag Lt from the portion stored in the long term filter state buffer unit. In such a case, in most cases, data is extracted during a subframe period that overlaps with the combined excitation signal of the past two subframes. The final combined excitation signals ex (−1), ex (−2), ex (−) of the last several subframes stored in the long-term filter state buffer unit
3), fixed point positions fp (-1), f of ex (-4)
p (−2), fp (−3), and fp (−4) are often different, and the fixed-point positions of the internal variables in the long term lag selection unit 205 are determined based on only the average voice power R0. It is not reasonable to determine

The decimal point position selection unit 207 outputs the final combined excitation signal ex (−
1), fixed-point positions fp (-1), fp (-2), fp (-) of ex (-2), ex (-3), ex (-4)
3) In consideration of fp (-4), the fixed-point position of the internal variable in the long term lag selecting unit 205 is determined as follows.

FIG. 5 shows a table structure incorporated in the decimal point position selection unit 207. The decimal point position selection unit 207 accesses the table shown in FIG. 5A based on the average voice power R0 of the current subframe, and estimates the fixed point position fp (0) obtained for the combined excitation signal of the current subframe. Get. Next, the decimal point position selection unit 207 converts the fixed point positions fp (−1), fp (−2), and fp stored in the decimal point position buffer unit 206.
(−3), fp (−4) and the fixed point position FP representing the largest number among the fixed point positions fp (0) obtained for the current subframe are obtained. After that, the decimal point position selection unit 207 uses this fixed point position FP as shown in FIG.
By accessing the table shown in (b), the fixed-point position is determined for each of the internal variables in the long term lag selection unit 205.

By determining the fixed-point position as described above, the long-term excitation code vector c0t
When searching for (long term lag Lt), it is possible to prevent the accuracy from lowering.

In the first code selector 208, the second code selector 209, and the gain selector 210, as in the case of the target signal generator 204, the process of determining the fixed-point position of the internal variable based on the average voice power R0 is performed. (See FIG. 3). In the first code selection unit 208, the second code selection unit 209, and the gain selection unit 210, the processing itself performed after determining the fixed-point position of such an internal variable is performed by the conventional first code selection unit shown in FIG. 10
4, the second code selection unit 105 and the gain selection unit 106 are substantially the same, and thus description thereof is omitted.

According to the VSELP speech encoding apparatus of the above embodiment, the fixed-point positions of the internal variables used for the operation in the long-term lag selecting unit 205 are fixed for the stored several immediately preceding synthetic excitation signals. Decimal point position,
In addition, since the determination is made based on the average voice power of the current subframe, the accuracy is not affected by the difference in the fixed-point position of the past synthetic excitation signal, the encoding accuracy of the audio signal is increased, and decoding is performed. The quality of the sound signal thus obtained can be improved as compared with the conventional case.

Note that the long-term excitation code vector c
For each candidate of 0 (long term lag L), the fixed-point position of an internal variable is determined from the fixed-point position of the combined excitation signal of the past two sub-frames relating to the candidate and the average voice power R0 of the current sub-frame. It is also conceivable to decide.
However, in this case, it is not practical because it is necessary to review the internal variables for each candidate of the long-term excitation code vector c0 (long-term lag L). It is preferable that the fixed-point positions of the internal variables be determined uniformly for the candidates.

Further, according to the VSELP speech coding apparatus of the above embodiment, when the speech mean power of a certain sub-frame is determined as the average information value from the speech mean power of two frames, the quantization of the speech mean power is performed. Since the arithmetic average processing of the index by quantization is performed, the configuration for obtaining the average audio power of the subframe can be simplified, or the number of processing steps can be reduced (thus, the power consumption can be reduced). it can).

(B) VSELP Speech Decoding Apparatus Next, a VSELP speech decoding apparatus according to an embodiment corresponding to the VSELP speech encoding apparatus according to the above embodiment will be described in detail with reference to the drawings.

FIG. 6 shows a functional block configuration of the VSELP speech decoding apparatus of this embodiment. In FIG. 6, information relating to the broken lines among the input lines to the respective functional blocks indicates information transmitted to the decoding device, and information relating to the solid lines among the input lines to the respective functional blocks. Is information formed in the decoding device.

As shown in FIG. 6, the VSELP speech decoding apparatus of this embodiment has a long term lag selecting section 30.
1. Decimal point buffer 302, decimal point selector 3
03, first code selector 304, second code selector 30
5, gain selector 306, three amplifier circuits 307 to 30
9, an adder circuit 310, a synthesis filter unit 311, and a post filter unit 312.

In FIG. 6, of the information transmitted from the VSELP speech coding apparatus, the long term lag L is given to the long term lag selecting section 301, and the index I of the first excitation code vector is set to the first code selecting section. 304
And the index J of the second excitation code vector
Is given to the second code selection unit 305, the gain index G is given to the gain selection unit 306, and the average voice power R0 is calculated by the decimal point position selection unit 303 and the first code selection unit 3
04, a second code selector 305, a gain selector 306,
The LPC coefficient kj is provided to the synthesis filter unit 311 and the post filter unit 312, and the LPC coefficient kj is provided to the gain selection unit 306, the synthesis filter unit 311 and the post filter unit 312.

The long-term lag selecting section 301 is provided with the synthesized excitation signal ex determined for each sub-frame,
The internal excitation signal e is stored in an internal long-term filter state buffer.
The information of the fixed-point position fp related to x is given to the decimal-point position buffer unit 302 and stored. That is, the contents stored in the long-term filter state buffer unit and the decimal point position buffer unit 302 before starting the processing of the current subframe are the same as those of the corresponding buffer unit in the encoding device.

The decimal point position selection unit 303 obtains an estimated value of the fixed point position fp (0) obtained for the combined excitation signal of the current subframe based on the average voice power R0 of the current subframe. The composite excitation signals ex (-1), ex (-2), ex (-3), ex (-
4) Fixed-point positions fp (-1), fp (-2), f
p (−3), fp (−4) and the fixed point position FP representing the largest number among the fixed point positions fp (0) obtained for the current subframe are obtained.
Thus, the fixed-point position of the internal variable in the long term lag selecting unit 301 is determined.

The long term lag selector 30 of the decoding device
In No. 1, since the search for the optimum long-term excitation code vector c0 is unnecessary, the variable whose fixed point is switched is the long-term excitation code vector c0 to be output.
Is itself.

When the fixed-point position with respect to the current subframe is determined, the long-term lag selecting section 301 extracts the long-term excitation code vector c0 from the long-term filter state buffer section based on the given long-term lag L. Then, the signal is output to the gain selection unit 306 and the amplification circuit 307.

First code selection section 304, second code selection section 305, gain selection section 306, synthesis filter section 31
1. The post-filter unit 312 determines the fixed-point position of an internal variable to be used for processing based on the average audio power R0 given from the encoding apparatus before executing the original processing assigned to the self-block. (See FIG. 3).
Note that the first code selection unit 30 in the speech decoding device
4. Second code selector 305 and gain selector 306
Does not require an optimum value search process, so the number of types of internal variables whose fixed-point positions are determined is smaller than that of the corresponding function block in the encoding device.

The first code selection unit 304 stores the same basis vector v1 as that of the first code selection unit 208 in the encoding device. When the fixed-point position is determined, the first vector selection unit 304 receives this basis vector and the received first base vector. The optimum first excitation code vector c1 is decoded from the index I for the excitation code vector and output to the amplification circuit 308 and the gain selection unit 306.

The second code selection unit 305 stores the same basis vector v2 as that of the second code selection unit 209 in the encoding device. When the fixed-point position is determined, the second vector selection unit 305 and the received second vector are received. The optimum second excitation code vector c0 is decoded from the index J of the excitation code vector and output to the amplification circuit 309 and the gain selection unit 306.

The gain selection unit 306 has the same gain parameters GS,
When the fixed-point position is determined, the optimum set of gain parameters GS, P0, P1 is extracted from the received gain index G, and the obtained long-term excitation code vector c0 is stored. , A first excitation code vector c1, a second excitation code vector c2,
The optimum gains β, γ1 and γ2 are decoded from the received LPC coefficient kj and the average audio power R0 and output to the corresponding amplifier circuits 307, 308 and 309, respectively.

The long-term excitation code vector c0, the first excitation code vector c1,
The second excitation code vector c2 is stored in each of the amplifier circuits 307, 3
08, 309, the corresponding gain β, γ1, γ2
Is controlled based on
To obtain a combined excitation signal ex for the current subframe. This combined excitation signal ex is provided to the combining filter unit 311 and the long term lag selecting unit 301.

The synthesis filter unit 311 sets the LPC coefficient kj received in the determined fixed-point position expression to define the filter characteristics, filters the input synthesized excitation signal ex with the filter characteristics, and converts the audio signal into an audio signal. Reproduce. The noise component of the reproduced audio signal obtained in this way may be large.

The post-filter unit 312 sets the LPC coefficient kj received in the determined fixed-point position expression to define the filter characteristics, filters the audio signal reproduced by the synthesis filter unit 311 with the filter characteristics, and removes the noise component. Is obtained and output.

According to the VSELP speech decoding apparatus of the above embodiment, the fixed-point positions of the internal variables used for the operation in the long term lag selecting section 301 are fixed for the stored several immediately preceding synthetic excitation signals. Decimal point position,
In addition, since the determination is made based on the average voice power of the current subframe, the accuracy is not affected by the difference in the fixed-point position of the past synthetic excitation signal, the encoding accuracy of the audio signal is increased, and decoding is performed. The quality of the sound signal thus obtained can be improved as compared with the conventional case. Needless to say, the same effect can be obtained by adopting the same fixed point position determining method as that of the encoding device.

(C) Other Embodiments In the description of the above embodiments, other embodiments have been mentioned. In addition to the above, the following other embodiments can be mentioned.

The features of the above embodiment are roughly two. The first feature relates to a method for interpolating the average voice power.
That is, the arithmetic mean of the index is used to obtain the average information from the two audio average powers. The second feature relates to a fixed-point representation method when a combined excitation signal of a past subframe is used as an excitation code vector component of a current subframe. That is, the fixed-point position is determined. However, the device may be configured to include only one of the features.

In the above embodiment, the present invention is applied to an apparatus according to the forward-type VSELP speech coding scheme. However, the present invention may be applied to an apparatus according to the backward-type VSELP speech coding scheme. CELP
The present invention may be applied to a device that conforms to a speech coding method.

For example, an apparatus according to a general CELP speech coding system also has an adaptive excitation code vector selection unit (corresponding to the long term lag selection unit in the embodiment) for storing and using past synthesized excitation signals. In such an apparatus, the second feature can be applied. In this case, a device that does not decompose a frame into subframes may be used. VSELP speech coding system
Is a type of CELP speech coding method,
The name of the component that stores and uses the synthesized excitation signal is
In the range of the
Is showing.

Further, for example, even if the apparatus does not use the average information of the two audio average powers for switching the fixed-point representation, it is necessary to obtain the average information of the two audio average powers. Then, the first characteristic configuration can be applied.

[0103]

As described above, according to the speech encoding apparatus and speech decoding apparatus of the present invention, each of the synthesized excitation signals of the past several frames stored in the adaptive excitation code vector selection unit is used. The adaptive excitation code vector selection unit uses a decimal point position buffer unit for storing fixed point position information, a plurality of fixed point position information stored in the decimal point position buffer unit, and a speech average power for the current frame. Since a decimal point position selector for determining the fixed point position of the variable is provided, even in the adaptive excitation code vector selector that stores the past synthesized excitation signal, it is possible to realize an improvement in accuracy by switching the fixed point position. become able to.

According to another speech coding apparatus of the present invention, the speech power interpolator includes at least speech average power quantizing means, and obtains speech average power information for each sub-frame. The audio average power of the two frames is treated as a quantized index, and the geometric average information of the audio average power of the two frames is obtained by the arithmetic mean of the index numbers, and the index obtained for each subframe is output or obtained. Since the average voice power obtained by returning the index obtained through the inverse quantization means is output, the average voice power information for each subframe can be obtained by a simple configuration or simple processing.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of a VSELP speech encoding device according to an embodiment.

FIG. 2 is a functional block diagram of a conventional VSELP speech coding apparatus.

FIG. 3 is an explanatory diagram of a general switching method of fixed-point representation.

FIG. 4 is an explanatory diagram (part 1) of a method for switching a fixed-point expression regarding a long-term lag selector according to the embodiment.

FIG. 5 is a diagram (part 2) illustrating a method of switching the fixed-point representation regarding the long-term lag selector according to the embodiment.

FIG. 6 is a functional block diagram of the VSELP speech decoding device according to the embodiment;

[Explanation of symbols]

 201: audio power calculation unit, 202: audio power interpolation unit, 202a: audio average power quantization table, 203: LPC analysis unit, 204: target signal creation unit, 205, 301: long term lag selection unit, 206, 302 ... Decimal point position buffer section, 207, 303... Decimal point position selecting section, 208, 304... First code selecting section, 209, 305. ... Amplifier circuits 214 and 310... Addition circuits 214 and 311... Synthesis filter section 312.

Continuation of the front page (72) Inventor Katsutoshi Ito 1-7-12 Toranomon, Minato-ku, Tokyo Oki Electric Industry Co., Ltd. (56) References JP-A-4-190399 (JP, A) JP-A-5-190 165498 (JP, A) JP-A-58-63998 (JP, A) JP-A-63-261925 (JP, A) JP-A-4-107600 (JP, A) JP-A-4-270399 (JP, A) JP-A-5-6199 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 19/00-19/14 H03M 7/30 H04B 14/04

Claims (6)

(57) [Claims] 1. A speech coding apparatus according to a CELP speech coding system for parameter coding a speech signal by digital processing with a finite word length, wherein a synthetic excitation signal of several past frames is stored. outputting a portion of the signal sequence as adaptive excitation code vector component, in speech encoding apparatus having an adaptive excitation code vector selection unit capable of switching fixed-point position of the internal variable, to the adaptive excitation code vector selection unit For each of the stored excitation signals of the past several frames, a fixed-point position buffer for storing fixed-point position information, a plurality of fixed-point position information stored in the fixed-point position buffer, and a voice average power of the frame, fixed-internal variables in the adaptive excitation code vector selection unit Speech coding apparatus is characterized in that a decimal point position selection section which determines a point location. 2. A speech encoding apparatus according to a CELP speech encoding method for encoding a speech signal by digital processing with a finite word length according to a CELP speech encoding method, comprising: a speech power calculation unit for calculating a speech average power of each frame of the speech signal; A voice power interpolator for obtaining voice average power information for each sub-frame obtained by equally dividing one frame into several frames from the voice average power of two successive frames. The unit includes at least voice average power quantization means, and when obtaining voice average power information for each subframe, treats the voice average power for each frame with a quantized index and calculates the voice average power of two frames. When calculating the geometric mean information, the arithmetic mean of the index is calculated, and the index obtained for each subframe is calculated. Force and, or, the speech coding apparatus and outputs a voice average power back through the inverse quantization means indexes obtained. 3. A speech encoding apparatus according to a CELP speech encoding method for encoding a speech signal by digital processing with a finite word length according to a CELP speech encoding method, comprising: a speech power calculation unit for calculating a speech average power of each frame of the speech signal. A voice power interpolator for obtaining voice average power information for each subframe obtained by equally dividing one frame into several frames from the voice average power of two successive frames, and a synthetic excitation signal of several past subframes Is stored, and a part of the signal sequence is output as an adaptive excitation code vector component. The power interpolation unit includes at least a voice average power quantizing unit, and obtains voice average power information for each sub-frame. The audio average power of each frame is treated as a quantized index, and when calculating geometric average information of the audio average power of two frames, the arithmetic average of the indexes is calculated, and the index determined for each subframe is output. or outputs the audio average power index back through the inverse quantization unit obtained, each of the synthetic excitation signal of the past several sub frames stored in said adaptive excitation code vector selection unit The adaptive excitation code vector selection is performed based on a decimal point position buffer unit for storing fixed point position information, a plurality of fixed point position information stored in the decimal point position buffer unit, and a speech average power for the current subframe. And a decimal point position selecting unit for determining a fixed point position of an internal variable in the unit. Speech coding apparatus for. 4. The speech encoding apparatus according to claim 1, wherein said CELP speech encoding scheme is one of the forward type VSELP speech encoding schemes. 5. A C code corresponding to the speech coding apparatus according to claim 1.
An audio decoding device according to an ELP audio encoding method,
Save the combined excitation signals of several past frames,
Outputting a portion of the signal sequence as adaptive excitation code vector component, the speech decoder with adaptive excitation code vector selection unit capable of switching fixed-point position of the internal variable, to the adaptive excitation code vector selection unit For each of the stored excitation signals of the past several frames, a fixed-point position buffer for storing fixed-point position information, a plurality of fixed-point position information stored in the fixed-point position buffer, and a voice average power of the frame, the speech decoding apparatus is characterized in that a decimal point position selection section which determines a fixed-point position of the internal variable in the adaptive excitation code vector selection unit. 6. The speech decoding apparatus according to claim 5, wherein said CELP speech coding system is one of the forward type VSELP speech coding systems.