JP2010276848A - Bidirectional predictive coding device and method, bidirectional predictive decoding device and method, and program and recording medium therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a predictive coding device etc. capable of improving efficiency of coding of a sound signal, in particular, the efficiency of coding when a frame length is short. <P>SOLUTION: A prediction coefficient calculating section 120 performs linear predictive analysis on an input signal constituted of sample columns of time series by a frame unit constituted of a plurality of samples, to calculate a prediction coefficient, and a prediction coefficient coding section 130 encodes the prediction coefficient to output a prediction coefficient code. A linear prediction section 140, to which the input signal and the prediction coefficient corresponding to the prediction coefficient code are input, selects one of prediction direction of a forward direction or the reverse direction in terms of time within the frame, and outputs a prediction error column calculated by the prediction coefficient corresponding to the input signal and the prediction coefficient code by predicting the selected prediction direction, and information indicating the selected prediction direction. A prediction error encoding section 150 encodes the prediction error column and outputs the prediction error code column. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a bidirectional predictive encoding apparatus, a bidirectional predictive decoding apparatus, a method thereof, a program thereof, and a recording medium thereof, which compress and encode and decode an audio signal without distortion.

  Non-patent document 1 discloses an encoding technique (reversible encoding that does not allow distortion) that can compress and expand information such as sound and images without distortion for linear PCM signals that have numerical values as amplitudes. Various methods have been devised.

  On the other hand, not long-distance transmission of telephones or voice transmission for VoIP, but ITU-T G. Logarithmic approximate companding PCM using 8 bits per sample, standardized as 711, is used. FIG. 18 shows an example of conversion characteristics from a linear PCM sample value to a logarithm approximate companding PCM sample value. 711 μ-law. For such a logarithmic approximate companded PCM signal, for example, the logarithmic approximate companded PCM signal is converted into a linear PCM signal, linear prediction is performed in the linear PCM domain, and the predicted value is logarithmic companded to obtain a residual. Thus, reversible encoding that does not allow distortion is possible.

  When predictive coding is performed by these conventional methods, as schematically shown in FIG. 3 (a), in the frame composed of a plurality of samples of the input signal, each input signal sample is temporally past. A prediction error sample is calculated by obtaining a weighted sum of the samples of the input signal as a predicted value of the current sample and subtracting the predicted value from the value of the input signal sample. Note that FIG. 3A shows that one sample of prediction error at a certain time point is obtained by subtracting, for example, a prediction value that is a weighted sum of past input signals of three samples from the input sample value at that time point. It is a schematic one.

Mat Hans, Ronald W. Schafer, "Lossless Compression of Digital Audio," IEEE Signal Processing Magazine, July 2001, pp, 21-32

As VoIP systems become popular in place of ordinary telephones, their transmission capacity increases. In VoIP, due to a request to reduce the delay time, the frame length as a unit of compression is short, and for example, there are cases where it is 40 samples per frame. However, if the frame length is short, the encoding efficiency (compression efficiency) decreases.
An object of the present invention is to realize a predictive encoding device, a predictive decoding device, and the like that can improve the efficiency of encoding a speech signal, particularly when the frame length is short.

The bidirectional predictive coding apparatus of the present invention includes a prediction coefficient calculation unit, a prediction coefficient coding unit, a linear prediction unit, and a prediction error coding unit.
The prediction coefficient calculation unit obtains a prediction coefficient by performing linear prediction analysis on an input signal composed of a time-series sample sequence in units of frames including a plurality of samples, and the prediction coefficient encoding unit encodes the prediction coefficient to perform prediction. Output coefficient code.

  The linear prediction unit receives the input signal and a prediction coefficient corresponding to the prediction coefficient code, selects a prediction direction that is temporally forward or backward in the frame, and selects the selected prediction direction. The prediction error sequence obtained from the input signal and the prediction coefficient corresponding to the prediction coefficient code and the information indicating the selected prediction direction are output. The prediction error encoding unit encodes the prediction error sequence and outputs a prediction error code sequence.

The bidirectional predictive decoding apparatus of the present invention includes a prediction coefficient decoding unit, a prediction error sequence decoding unit, and a linear prediction decoding unit.
The prediction coefficient decoding unit decodes the prediction coefficient from the prediction coefficient code, and the prediction error sequence decoding unit decodes the prediction error sequence from the prediction error code sequence.
The linear prediction decoding unit restores the original signal by prediction in the prediction direction based on the prediction coefficient, the prediction error string, and information indicating the prediction direction of the prediction error string.

  The waveform of the prediction error generated differs depending on the direction of prediction. When the waveform is encoded without distortion, a prediction direction with a small amount of code can be selected according to the present invention, and even if 1 bit for transmitting the prediction direction is included as a whole The amount of codes can be reduced.

The figure which shows the function structural example of the bi-directional predictive coding apparatus 100,200. The figure which shows the example of a processing flow of the bi-directional predictive coding apparatus. (a) is a figure which shows the prediction image of a forward direction, (b) is a figure which shows the prediction image of a reverse direction. The figure which shows the function structural example of the bi-directional predictive decoding apparatus 600 and 600a. The figure which shows the example of a processing flow of the bi-directional predictive decoding apparatus 600. The figure which shows the function structural example of the bi-directional predictive coding apparatus 100a. The figure which shows the function structural example of the bi-directional predictive coding apparatus 100b. The figure which shows the function structural example of the bi-directional predictive decoding apparatus 600b. The figure which shows the function structural example of the bi-directional predictive coding apparatus 200a. The figure which shows the function structural example of the bidirectional | two-way prediction decoding apparatus 700. The figure which shows the example of the code allocation in the case of carrying out variable length coding integrally with prediction order information and prediction direction information. The figure which shows another example of the code allocation in the case of carrying out variable length coding integrally with prediction order information and prediction direction information. The figure which shows the function structural example of the bi-directional predictive coding apparatus 300. The figure which shows the function structural example of the bidirectional | two-way prediction decoding apparatus 800. The figure which shows the function structural example of the bi-directional predictive coding apparatus 400. The figure which shows the function structural example of the bi-directional predictive coding apparatus 500. The figure which shows the function structural example of the bidirectional | two-way prediction decoding apparatus 900. The figure which shows the example of the conversion characteristic of linear PCM and logarithm approximation companding PCM.

Hereinafter, embodiments of the present invention will be described in detail.

[Encoder]
FIG. 1 shows a functional configuration example of the bidirectional predictive coding apparatus 100 of the present invention, and FIG. 2 shows a processing flow example thereof. The bidirectional predictive encoding device 100 includes a frame dividing unit 110, a prediction coefficient calculating unit 120, a prediction coefficient encoding unit 130, a linear prediction unit 140, a prediction error encoding unit 150, and a multiplexing unit 160.
A time series sample x (i) (where i is a number indicating the time order of the time series samples) is input to the frame division unit 110, and a signal sequence X = {x (1), x (2),..., x (N)} are output. Here, N is the number of samples in one frame.

The prediction coefficient calculation unit 120 includes a prediction analysis unit 121 and a prediction coefficient quantization unit 122. The prediction analysis unit 121 receives the signal sequence X and performs prediction analysis to perform prediction analysis with a predetermined order P of the prediction coefficient K = {k (1), k (2),..., K ( P)}. As a form of prediction analysis, there are short-term prediction having a coefficient from the first order to several orders in succession and long-term prediction having a coefficient between samples separated by a pitch period, but any form may be used. . Further, the order P may be determined by another means (not shown) without being predetermined. As a method for determining the order P for each frame, for example, there is a method described in JP-A-2009-69309. The prediction coefficient quantization means 122 quantizes the prediction coefficient K and outputs quantized prediction coefficients K _q = {k _q (1), k _q (2),..., K _q (P)} (S1). ). Note that the amount of information at the time of encoding can be suppressed by converting the prediction coefficient into a PARCOR (Partial Autocorrelation) coefficient prior to quantization.

The prediction coefficient encoding unit 130 receives the quantized prediction coefficient K _q , encodes the prediction coefficient code C _k = {c _k (1), c _k (2),..., C _k (P )} Is output (S2). When encoding, variable length encoding is performed by assigning a short code to a value with a high appearance frequency and assigning a long code to a value with a low appearance frequency, so that the code length can be shortened. Can be further enhanced.
The linear prediction unit 140 receives the signal sequence X and the quantized prediction coefficient K _q , selects either the forward direction or the reverse direction in time within the frame, and performs the signal based on the prediction of the selected prediction direction. A prediction error sequence E = {e (1), e (2),..., E (N)} obtained from the sequence X and the quantized prediction coefficient _Kq, and information s indicating the selected prediction direction Output (S3).

  As described above, when obtaining a predicted value sequence, as shown in FIG. 3 (a), the current sample is obtained by calculating a weighted sum of past samples in time within a frame consisting of a plurality of samples. (Forward prediction). However, as shown in FIG. 3B, it is also possible to predict the current sample by calculating a weighted sum of future samples in the frame (prediction in the reverse direction). The prediction coefficient is obtained from the correlation function of the sample in the frame by a normal linear prediction analysis, and this correlation function has no difference before and after time. As a result, the prediction coefficient in the forward direction and the prediction coefficient in the reverse direction are the same, but even if the coefficients are exactly the same, both prediction error sequences are different, and the code when encoded (particularly when variable-length encoding is performed) The amount is also different. Therefore, the code amount to be transmitted to the decoding side can be reduced by selecting the prediction error sequence in the prediction direction with the smaller code amount.

For example, as shown in FIG. 1, the linear prediction unit 140 can include a prediction unit 141, a subtraction unit 142, and a prediction direction selection unit 143. The prediction unit 141 uses the signal sequence X and the quantized prediction coefficient K _q to predict the forward direction in the frame according to the expression (1) in the time direction and the prediction in the time direction according to the expression (2). Predicted value sequence X _pf = {x _pf (1) , x _pf (2), ..., x _pf (N) } And the backward predicted value sequence X _pb = {x _pb (1), x _pb (2),..., X _pb (N) }

In addition, when i-j of x (i-j) used in Expression (1) is outside the range of 1 to N, for example, prediction by Expression (1) is performed assuming that the value of x (i-j) is 0. Alternatively, the prediction is performed by setting the addition range of Σ in the equation (1) not to P but to the maximum value where i-j is not outside the range of 1 to N. This prediction method is specifically described in Japanese Patent No. 3871672. The same applies to i + j of x (i + j) used in Equation (2).

Subtracter means 142, the signal sequence X forward prediction value sequence X _pf the corresponding order is subtracted for each sample direction prediction error sequence from _{E f = {e f (1} ), e f (2), ···, e _f (N)} and the backward prediction value sequence X _pb subtracted from the signal sequence X for each corresponding sample, the backward prediction error sequence E _b = {e _b (1), e _b (2),. , e _b (N)} respectively. The prediction direction selection unit 143 encodes the code amount obtained by encoding the forward prediction error sequence E _f by a predetermined method and the code obtained by encoding the backward prediction error sequence E _b by a predetermined method. Are respectively selected, the prediction error sequence having the smaller code amount is selected and output, and information s indicating the prediction direction corresponding to the selected prediction error sequence E is output. As the information s indicating the prediction direction, for example, as 1-bit information, it is conceivable that 0 is output in the forward direction and 1 is output in the reverse direction. Note that it is more practical to use the estimated value without calculating the actual code amount for selection. As the estimated value, the sum of absolute values of the amplitudes of the respective sequences of the forward prediction error sequence E _f and the backward prediction error sequence E _b or the square sum of the amplitudes can be used. This is because the code amount tends to be approximately proportional to the sum of absolute values or the sum of squares of amplitudes. Also, when selecting the direction, instead of calculating all prediction errors for the entire frame, obtain a bidirectional prediction error halfway or for some samples and estimate the intermediate code amount in each direction. The prediction error may be calculated to the end only in the direction in which the code amount decreases. Since the direction in which the final code amount is likely to increase at the intermediate code amount stage is not calculated, the calculation amount can be reduced. For example, first, the prediction unit 141 includes M forward prediction values x _pf (1), x _pf (2),..., X _pf (M), where N> M, and M backward prediction values x. _pb (N-M + 1), _xpb (N-M + 2),..., _xpb (N) are obtained, and the subtracting means 142 _obtains M forward prediction errors e _f (1), e _f (2), ..., e _f (M) and M backward prediction errors e _b (N-M + 1), e _b (N-M + 2), ..., e _b (N) Ask for. The prediction direction selection unit 143 obtains the sum of absolute values of amplitudes or the sum of squares of amplitudes of M forward prediction errors and M backward prediction errors as an estimated value of the code amount, and the obtained estimated value is small. Information s indicating the prediction direction is output. Next, prediction values and prediction errors corresponding to NM samples other than the above for the selected prediction direction are obtained by the prediction means 141 and the subtraction means 142, respectively, and the selected prediction direction has already been obtained. A sum of the M prediction errors and the newly obtained NM prediction errors is output as a prediction error sequence E.

The prediction error encoding unit 150 encodes the prediction error sequence E to generate a prediction error code sequence C _e = {c _e (1), c _e (2),..., C _e (N)}. Output (S4). When encoding, variable length encoding is performed by assigning a short code to a value with a high appearance frequency and allocating a long code to a value with a low appearance frequency, thereby reducing the code length and improving the compression efficiency. Can be increased. When the prediction direction selection unit 143 is configured to obtain the actual code amount, the bidirectional prediction encoding apparatus 100 is configured not to include the prediction error encoding unit 150 and the prediction direction selection unit 143 has selected the prediction. and it outputs a code corresponding to the error row E as a prediction error code sequence C _e.
Finally, the multiplexing unit 160 multiplexes the prediction error code sequence C _e and information s and the prediction coefficient code C _k, and outputs a multiplexed signal M.

[Decoding device]
FIG. 4 shows a functional configuration example of the bidirectional predictive decoding apparatus 600 of the present invention, and FIG. The bi-directional predictive decoding device 600 includes a demultiplexing unit 610, a prediction coefficient decoding unit 620, a prediction error decoding unit 630, and a linear prediction decoding unit 640, and is opposed to the bi-directional predictive coding device 100 of the present invention shown in FIG. Used.

Demultiplexer 610 receives the multiplexed signal M transmitted from the encoding device, a prediction coefficient code C _k, and the prediction error code sequence C _e, and outputs the separated information s indicating a prediction direction.
Prediction coefficient decoding unit 620, a prediction coefficient code _{C k} is decoded into the quantized prediction coefficients _{K q} (S11), the prediction error decoding unit 630 decodes the prediction error code sequence _{C e} to the prediction error sequence E (S12) .

The linear prediction decoding unit 640 receives the quantized prediction coefficient _Kq , the prediction error sequence E, and the information s indicating the prediction direction. If the prediction direction obtained from the information s is the forward direction, i is from 1 to N. The original signal sequence X is obtained by sequentially generating the original signal x (i) by the forward prediction according to the formula (3) in this order. Also, if the prediction direction is the reverse direction, the original signal sequence X is obtained by sequentially generating the original signal x (i) by the prediction in the reverse direction according to Equation (4) in the order from i to N to 1. (S13).

Note that when ij of x (ij) used in the expression (3) is outside the range of 1 to N, for example, the expression (3) is performed assuming that the value of x (ij) is 0, or the expression ( The prediction is performed with the addition range of Σ in 3) not up to P, but up to the maximum value where ij is not outside the range of 1 to N. The same applies to i + j of x (i + j) used in equation (4).
By configuring the encoding device and the decoding device as described above, it is possible to adaptively select a prediction direction that can be encoded more efficiently and perform predictive encoding. Therefore, when the encoding efficiency, particularly when the frame length is short, Efficiency can be improved.

<Modification 1>
FIG. 6 shows a functional configuration example of a bidirectional predictive encoding device 100a which is a modification of the bidirectional predictive encoding device 100. The difference from the bi-directional predictive coding apparatus 100 is that a linear conversion unit 115 is inserted after the frame dividing unit 110, and a companding conversion unit 144 is inserted after the linear predictive coding unit 140 after the prediction unit 141. In addition, the linear predictive encoding unit 140a is replaced. Other configurations are the same.

When the input time-series sample x (i) is a companded signal such as a logarithmic approximate companded PCM signal, the linear PCM signal before companding can be obtained by using the bidirectional predictive encoding device 100 of FIG. Due to the non-linearity with respect to, the prediction error becomes large, and the coding efficiency decreases. This is particularly noticeable in samples with a large amplitude. Therefore, in the bidirectional predictive encoding device 100a of the present modification, the signal sequence X output in units of frames from the frame dividing unit 110 is linearly similar to the signal sequence before companding by the linear conversion unit 115 or a signal having a relationship close thereto. After conversion to the sequence Y, the signal sequence Y is used instead of the signal sequence X to perform processing in the prediction coefficient calculation unit 120 and the prediction unit 141. Then, the predicted value sequences Y _pf and Y _pb obtained by the predicting unit 141 are converted into signal sequences X _pf and X _{pb, which} are companded by the companding conversion unit 144. The conversion performed by the companding conversion unit 144 is an inverse conversion of the conversion in the linear conversion unit 115.

  Also, the bidirectional predictive decoding device 600a used opposite to the bidirectional predictive coding device 100a is the same as the bidirectional predictive decoding device 600 except that the linear predictive decoding unit 640 shown in FIG. 4 is replaced with the linear predictive decoding unit 640a. It is a configuration. The linear prediction decoding unit 640a generates the original signal sequence by sequentially generating x (i) according to Equation (5) when forward prediction is performed and according to Equation (6) when backward prediction is performed. Get X.

Here, x _pf (i) is a value obtained by companding y _pf (i) obtained by equation (7), and x _pb (i) is y _pb (i) obtained by equation (8). It is a value obtained by extension conversion. The companding conversion is a conversion for converting a companded signal sequence such as a logarithmic approximate companded PCM signal into a linear or close relationship with a linear PCM signal sequence.

First, y _pf (i) is obtained for i starting from 1 by equation (7), converted to x _pf (i) by the inverse transformation of the companding transformation, and x (i) by equation (5). Is converted into y (i) by the above-described companding conversion. This y (i) is used for the synthesis after the time point i + 1 in equation (7). Similarly, but in the reverse order in terms of time, they are synthesized by equations (6) and (8). That is, for i starting from N, y _pb (i) is first obtained by equation (8), converted to x _pb (i) by the inverse transformation of the companding transformation, and x (i) is obtained by equation (6). Is converted into y (i) by the above-described companding conversion. This y (i) is used for the synthesis before the time point i−1 in the equation (8). Note that y (ij) and y (i + j) are obtained by converting x (ij) and x (i + j) into values that are linear or close to the values before the companding, respectively. If ij of x (ij) is outside the range of 1 to N, for example, the value of x (ij) is assumed to be 0, and Expression (7) is performed, or the addition range of Σ in Expression (7) is The prediction is performed not to P but to the maximum value where ij does not fall outside the range of 1 to N. The same applies to i + j of x (i + j).

  In this way, by calculating the prediction coefficient and the prediction value sequence after converting the signal sequence before the companding into a linear or close relationship signal sequence, the prediction error can be further reduced and the encoding efficiency can be improved. be able to.

<Modification 2>
As described above, there are two types of prediction analysis: short-term prediction having coefficients ranging from the first order to several orders in succession and long-term prediction having coefficients between samples separated by a pitch period. In the second modification, a bidirectional predictive encoding device 100b that performs predictive encoding by combining short-term prediction and long-term prediction and an opposing bidirectional predictive decoding device 600b will be described.
FIG. 7 shows a functional configuration example of the bidirectional predictive encoding device 100b. The bi-directional predictive encoding device 100b includes a frame dividing unit 110, a prediction coefficient calculating unit 120b, a prediction coefficient encoding unit 130b, a linear prediction unit 140b, a prediction error encoding unit 150, and a multiplexing unit 160. The functions of the frame division unit 110, the prediction error encoding unit 150, and the multiplexing unit 160 are the same as those of the bidirectional predictive encoding device 100.

The prediction coefficient calculation unit 120b includes a prediction analysis unit 121b and a prediction coefficient quantization unit 122b. The prediction analysis unit 121b receives the signal sequence X in units of frames, performs short-term prediction analysis and long-term prediction analysis, respectively, and performs a short-term prediction coefficient K ₁ = {k ₁ (1) of a predetermined order P in units of frames. _{), k 1 (2),} ···, long-term prediction coefficients of a predetermined degree Q and _{k 1 (P)} K 2} = {k 2 (1), k 2 (2), ···, k 2 (Q)} and the delay amount T are calculated. Predictive coefficient quantization means 122b are short-term prediction coefficients K ₁ and long-term prediction coefficient K ₂ and the each quantized quantized short-term prediction coefficient _{K 1q = {k 1q (1} ), k 1q (2), ···, k _1q (P)}, a quantized long-term prediction coefficient K _2q = {k _2q (1), k _2q (2),..., k _2q (Q)} and a delay amount T are output.

The prediction coefficient encoding unit 130b encodes the quantized short-term prediction coefficient K _1q , the quantized long-term prediction coefficient K _2q, and the delay amount T, respectively, so that the short-term prediction coefficient code C _k1 = {c _k1 (1), c _k1 ( 2), ..., c _k1 (P)} and the long-term prediction coefficient code C _k2 = {c _k2 (1), c _k2 (2), ..., c _k2 (Q)} and the delay amount code C _T Is output.

The linear prediction unit 140b includes a first prediction unit 141b, a first subtraction unit 142b, a first prediction direction selection unit 143b, a second prediction unit 145, a second subtraction unit 146, and a second prediction direction selection unit 147. The first prediction unit 141b uses the signal sequence X and the quantized short-term prediction coefficient _K1q to make a short-term prediction in the forward direction according to Equation (9) and a short-term in the reverse direction according to Equation (10) in the frame. By the prediction, the forward short-term predicted value sequence X _1pf = {x _1pf (1), x _1pf (2),..., X _1pf (N) } And the reverse short-term predicted value sequence X _1pb = {x _1pb (1), x _1pb (2),..., X _1pb (N)}.

When ij of x (ij) used in equation (9) is outside the range of 1 to N, for example, prediction by equation (9) is performed assuming that the value of x (ij) is 0. Alternatively, the prediction is performed by setting the addition range of Σ in the equation (9) not to P but to the maximum value where ij does not fall outside the range of 1 to N. The same applies to i + j of x (i + j) used in Expression (10). The first subtractor 142 subtracts the forward short-term predicted value sequence X _1pf from the signal sequence X for each corresponding sample, and the forward short-term predicted error sequence E _1f = {e _1f (1), e _1f (2) _{,. ..} , E _1f (N)} and the reverse short-term prediction error sequence X _1pb subtracted from the signal sequence X for each corresponding sample, the reverse short-term prediction error sequence E _1b = {e _1b (1), e _1b ( 2), ..., e _1b (N)} are obtained. The first prediction direction selection unit 143b encodes the code amount obtained by encoding the forward short-term prediction error sequence E _1f by a predetermined method and the reverse short-term prediction error sequence E _1b by a predetermined method. The code amount of the obtained code is obtained, the smaller prediction error sequence is selected, and the short-term prediction error sequence E ₁ = {e ₁ (1), e ₁ (2),..., E ₁ ( and outputting the results as N)}, and outputs the information s ₁ indicating a prediction direction of the short-term prediction corresponding to the short-term prediction error string E _1. As the information s ₁ , for example, as 1-bit information, 0 is output in the forward direction, and 1 is output in the reverse direction. Note that it is more practical to use the estimated value without calculating the actual code amount for selection. As the estimated value, the sum of absolute values of all or part of the series of the forward short-term prediction error sequence E _1f and the reverse short-term prediction error sequence E _1b , or the sum of squares of the amplitudes can be used. The second prediction means 145 uses the short-term prediction error sequence E ₁ , the quantized long-term prediction coefficient K _2q, and the delay amount T to calculate the long-term prediction in the forward direction according to the equation (11) in the frame and the equation (12). By the long-term prediction in the reverse direction in time, the forward long-term prediction value sequence X _2pf = {x _2pf (1), x _2pf (2),..., X _2pf (N)} and the reverse long-term prediction value sequence X _2pb = {x _2pb (1), x _2pb (2),..., X _2pb (N)} is obtained.

When ijT of e ₁ (ijT) used in Expression (11) is outside the range of 1 to N, for example, prediction based on Expression (11) is performed assuming that the value of e ₁ (ijT) is 0. . The same applies to i + j + T of e ₁ (i + j + T) used in equation (12). The second subtraction means 146 subtracts the forward long-term prediction value sequence X _2pf from the short-term prediction error sequence E _{1 for} each corresponding sample, and the forward prediction error sequence E _2f = {e _2f (1), e _2f (2). , ..., and e _2f (N)}, the short-term prediction error string E ₁ from the opposite direction the long-term prediction value sequence X _2pb corresponding backward prediction is subtracted for each sample error sequence _{E 2b = {e 2b (1} ) , e _2b (2),..., e _2b (N)}, respectively. The second prediction direction selection means 147 obtains the code amount of the code obtained by encoding the forward prediction error sequence E _2f by a predetermined method and the reverse prediction error sequence E _2b by a predetermined method. The code amount of the generated code is calculated, the smaller prediction error sequence is selected and output as a prediction error sequence E combining short-term prediction and long-term prediction, and the prediction direction of the long-term prediction corresponding to the prediction error sequence E outputs information _{s 2} showing the. For the information s ₂ , for example, as 1-bit information, 0 is output in the forward direction and 1 is output in the reverse direction. In selection, it is more realistic to use the estimated value without calculating the actual code amount. As the estimated value, the sum of absolute values of all or a part of the respective sequences of the forward prediction error sequence E _2f and the backward prediction error sequence E _2b , or the sum of squares of the amplitudes can be used. Further, the average code amount can be reduced by assigning a variable length code to each of the four states based on the information s ₁ and s ₂ indicating the prediction direction for the short-term prediction and the long-term prediction. When the second prediction direction selection unit 147 is configured to obtain the actual code amount, the bidirectional prediction encoding device 100b is configured not to include the prediction error encoding unit 150, and the second prediction direction selection unit 147 includes the second prediction direction selection unit 147. and it outputs a code corresponding to the selected prediction error sequence E as a prediction error code sequence C _e.

FIG. 8 shows a functional configuration example of a bidirectional predictive decoding device 600b used opposite to the bidirectional predictive encoding device 100b. The bi-directional predictive decoding device 600b includes a demultiplexing unit 610, a prediction coefficient decoding unit 620b, a prediction error decoding unit 630, a second linear prediction decoding unit 680, and a first linear prediction decoding unit 640b. The functions of the demultiplexing unit 610 and the prediction error decoding unit 630 are the same as those of the bidirectional predictive decoding device 600.
Prediction coefficient decoding section 620b decodes the delay code C _T and short-term prediction coefficient code C _k1 and the long-term prediction coefficient code C _k2, quantized short-term prediction coefficient K _1q and the quantized long-term prediction coefficient K _2q and the delay amount T Is output.

The second linear prediction decoding unit 680 receives the prediction error sequence E, the quantized long-term prediction coefficient K _2q , the delay amount T, and the information s ₂ indicating the prediction direction of the long-term prediction, and the prediction direction obtained from the information s ₂ is If the forward direction, the short-term prediction error sequence E ₁ by the forward prediction is generated according to the formula (13) in the order from i to 1 to N, and if i is the backward direction, the formula (14 ) by generating a short-term prediction error string E ₁ by backward prediction.

When ijT of e ₁ (ijT) used in Expression (13) is outside the range of 1 to N, for example, prediction by Expression (13) is performed assuming that the value of e ₁ (ijT) is 0. . The same applies to i + j + T of e ₁ (i + j + T) used in equation (14).

First linear prediction decoding unit 640b includes a short-term prediction error string E ₁ generated by the second linear prediction decoding unit 680, the quantized short-term prediction coefficient K _1q, and information s ₁ indicating a prediction direction of the short-term prediction If the prediction direction obtained from the information s ₁ is the forward direction, the original signal x (i) based on the forward prediction is sequentially generated by the formula (15) in the order from 1 to N, and the prediction direction If i is in the reverse direction, the original signal sequence X is obtained by sequentially generating the signal x (i) by the prediction in the reverse direction according to the equation (16) in the order from i to N to 1.

  Note that when i-j of x (i−j) used in Expression (15) is outside the range of 1 to N, for example, prediction by Expression (15) is performed assuming that the value of x (i−j) is 0. Alternatively, the prediction is performed by setting the addition range of Σ in Expression (15) not to P but to the maximum value where i-j is not outside the range of 1 to N. The same applies to i + j of x (i + j) used in Expression (16).

Note that when either the short-term prediction or the long-term prediction is selectively executed instead of performing both the short-term prediction and the long-term prediction on the encoding device side, either the short-term prediction or the long-term prediction is selected. 1-bit auxiliary information indicating this is output from the encoding device and transmitted to the decoding device, and the decoding device may select either short-term prediction or long-term prediction according to this auxiliary information and perform decoding.
As described above, by performing bidirectional prediction by combining short-term prediction and long-term prediction, the prediction error can be further reduced, so that the encoding efficiency can be further increased.

  In the bi-directional predictive encoding device 100 of the first embodiment, bi-directional prediction is performed in all cases, but the average processing amount increases when bi-directional prediction errors are obtained in all cases. As a method for suppressing this, for example, the following two methods can be considered.

  First, the bidirectional prediction error calculation is limited to a case where the calculation value is large. Specifically, the case where the effect of prediction is large, that is, the case where the prediction coefficient or the PARCOR coefficient is large can be considered. For example, a bidirectional prediction error is obtained only when the first-order PARCOR coefficient after inverse quantization is equal to or greater than a threshold value, and for other cases, for example, only the forward direction is obtained.

Second, the calculation of the bidirectional prediction error is limited to a case where the calculation effort is small. In predictive coding, the prediction order may be adaptively selected to reduce the sum of the information amount of prediction coefficients and the information amount of prediction errors. The bidirectional prediction of the present invention is effective regardless of the order, but the amount of calculation for obtaining the prediction error is almost twice that in the case of the unidirectional prediction. Moreover, since the number of coefficients increases as the predicted order increases, the amount of calculation increases greatly. Therefore, in order to obtain an effect of bi-directional prediction while suppressing an increase in the total amount of calculation, bi-directional prediction is performed only when the prediction order is equal to or less than a certain value. To make a prediction.
In any case, if the prediction direction in the case of prediction in one direction is determined in advance, the information s for transmitting the selected prediction direction may be transmitted only when bidirectional prediction is performed.

FIG. 1 shows a functional configuration example of a bidirectional predictive coding apparatus 200 of the present invention that embodies the first method. The bidirectional predictive encoding device 200 has a configuration in which the predicting unit 141 of the bidirectional predictive encoding device 100 is replaced with a predicting unit 241.
When the value of the primary prediction coefficient k (1) (primary quantized prediction coefficient k _q (1)) obtained by the prediction analysis means 121 is within a preset range, the prediction means 241 Processing similar to that of the prediction unit 141 is performed. On the other hand, when the value is out of the range, the predicted value sequence _Xp is obtained by prediction in any one direction set in advance, and thereafter, the process is performed only in the one direction. In this case, transmission / reception of the information s indicating the prediction direction is not necessary. This is because the predetermined one-way processing is performed on the encoding side when the value of the quantization prediction coefficient K _q received by the linear prediction decoding unit 640 on the decoding side is outside the predetermined range. This is because the decoding side can recognize this. That is, the quantized prediction coefficient _Kq also serves as information s indicating the prediction direction.

  FIG. 9 shows a functional configuration example of the bidirectional predictive encoding device 200a according to the present invention that embodies the second method. The bi-directional predictive encoding device 200a includes a frame dividing unit 110, a prediction coefficient calculating unit 220a, a prediction coefficient encoding unit 130, a linear prediction unit 240a, a prediction error encoding unit 150, and a multiplexing unit 160. The prediction coefficient calculation unit 220a includes a prediction analysis unit 221a and a prediction coefficient quantization unit 122, and the linear prediction unit 240a includes a prediction unit 241a, a subtraction unit 142, and a prediction direction selection unit 143. That is, the configuration is the same as that of the bidirectional predictive encoding device 100 except for the prediction analysis unit 221a and the prediction unit 241a.

The prediction analysis unit 221a adaptively determines the degree P of the prediction coefficient for each frame, and outputs it together with the prediction coefficient K. Further, since the order P of the prediction coefficient is also necessary for the prediction process in the decoding apparatus, the multiplexing unit 160 multiplexes it with other information and transmits it. When the order P is transmitted, a bidirectional predictive decoding apparatus 700 as shown in FIG. 10 is provided oppositely, and the order P is used for the prediction process in the linear predictive decoding unit 740. The prediction unit 241a performs the same processing as the prediction unit 141 when the order P of the prediction coefficient is within a preset range. On the other hand, when the value of the order P of the prediction coefficient is outside the preset range, the prediction value sequence _Xp is obtained by prediction in any one direction set in advance, and thereafter, only the one direction is processed. In this case, transmission / reception of the information s indicating the prediction direction is not necessary. This is because the order P of the prediction coefficient received by the linear prediction decoding unit 740 on the decoding side is out of the preset range, and that the preset one-way processing has been performed on the encoding side. This is because it can be recognized on the side. That is, the order P of the prediction coefficient also serves as information s indicating the prediction direction. Note that the prediction direction determination parameter does not necessarily have to be the order of the prediction coefficient, and a parameter applicable to the present invention can be selected as appropriate.
As described above, it is possible to prevent an increase in the average processing amount by restricting bidirectional prediction.

  When calculating the prediction coefficient in the prediction analysis means of the encoding apparatus, when the prediction coefficient order P is adaptively determined for each frame, for example, in the configuration example of FIGS. 9 and 10, the prediction order P and the prediction value sequence The information s indicating the prediction direction is handled individually, but the average code length can be reduced by handling the information s as a variable length code integrally. For example, as shown in FIG. 11, the prediction order P is 2 bits (selected from four types of primary, secondary, tertiary, and quadratic), and information s that conveys the prediction direction is 1 bit (0: forward direction, 1 : In the reverse direction), 3 bits are allocated in all cases. However, the average code length can be reduced by assigning variable length codes according to the frequency of each combination. The variable length allocation shown in FIG. 11 is an allocation example in which only the forward direction is used in the case of high-order prediction and the code length is also short. Further, the variable length allocation shown in FIG. 12 is an example of allocation when the frequencies of the orders 00 (primary) and 11 (quaternary) are high, and in this case, the frequency of selecting the reverse direction is higher than the forward direction.

  FIG. 13 shows an example of a functional configuration of a bidirectional predictive encoding device 300 that performs variable-length encoding, and FIG. 14 shows an example of a functional configuration of a bidirectional predictive decoding device 800 that is used opposite to the bidirectional predictive encoding device 300. . FIG. 13 is a configuration example based on the configuration of FIG. 9. P and s are not directly input to the multiplexing unit 160 but are input to the variable length encoding unit 370 and converted to the variable length code Ps. Then, it is input to the multiplexing unit 160 (S5 in the processing flow of FIG. 2). FIG. 14 is a configuration example based on the configuration of FIG. 10. The variable length code Ps output from the demultiplexing unit 610 is decoded into P and s by the variable length code decoding unit 805, and then linear. The data is input to the prediction unit 740 (S14 in the processing flow of FIG. 5).

  Thus, the average code length can be reduced by integrally variable-length encoding the prediction order P and the information s indicating the prediction direction of the prediction value sequence.

The fourth embodiment has a configuration in which forward / reverse determination is performed using a byproduct generated when a prediction coefficient is obtained by a sequential analysis method.
FIG. 15 shows a functional configuration example of the bidirectional predictive encoding device 400 of the present invention. The bi-directional predictive coding apparatus 400 includes a frame dividing unit 110, a prediction coefficient calculating unit 420, a prediction coefficient coding unit 130, a linear prediction unit 440, a prediction error coding unit 150, and a multiplexing unit 160. Further, the prediction coefficient calculation unit 420 includes a prediction analysis unit 421 and a prediction coefficient quantization unit 122, and the linear prediction unit 440 includes a prediction method determination unit 448, a prediction unit 441, and a subtraction unit 142. In other words, except for the absence of the prediction direction selection unit 143, the configuration other than the prediction analysis unit 421, the prediction method determination unit 448, and the prediction unit 441 is the same as that of the bidirectional predictive encoding device 100.

Prediction analysis unit 421 calculates by successive analysis by forward prediction and backward prediction a prediction coefficient K, and outputs with forward prediction error p _f and backward prediction error p _b which is obtained in the course of the sequential analysis. Examples of the sequential analysis method include a BURG method and a PARCOR lattice method.

Predictive method determination means 448, when the forward prediction error p _f is backward prediction error p _b is smaller than outputs information s indicating that prediction direction is forward, forward prediction error p _f is backward prediction error p When it is larger than _b, information s indicating that the prediction direction is the reverse direction is output. Incidentally, the magnitude relationship between the forward prediction error p _f and backward prediction error p _b, for example, compares the average amplitude or root-mean-square amplitude of the absolute value of the prediction error.

Then, the prediction unit 441 obtains a prediction value sequence X _p by prediction in the prediction direction indicated by the information s, and the subtraction unit 142 obtains a prediction error sequence E obtained by subtracting the prediction value sequence X _p from the signal sequence X to obtain a prediction error code. The encoding unit 150 encodes the prediction error sequence _E to generate and output an encoded prediction error sequence Ce.

BURG method or PARCOR grid method final stage or indeed amplitude of the magnitude of corresponding to the selected prediction order forward prediction error p _f and backward prediction error p _b in sequential analysis process, such as, the present application of the forward There is a large correlation between the magnitudes of the respective amplitudes of the prediction error string E _f and the prediction error string E _{b in} the reverse direction. That is, by comparing the forward prediction error p _f obtained in the course of the sequential analysis method and backward prediction error p _b, the larger the forward prediction errors, by performing a backward prediction in determining the predicted value sequence, forward There is a high possibility that the error will be smaller than when the prediction is performed. On the other hand, if the backward prediction error is small, if the prediction in the forward direction is performed when the prediction value sequence is obtained, there is a high possibility that the error is smaller than that in the case of performing the prediction in the backward direction. Therefore, the prediction error can be reduced on average without performing bidirectional prediction in the prediction means.
As described above, in this embodiment, the amount of calculation in the encoding device can be reduced by determining the prediction direction using the by-product generated in the step of obtaining the prediction coefficient.

<Modification>
If, when the amplitude of the ratio of the size of the forward prediction error p _f and backward prediction error p _b is close to 1, the prediction method determining unit 448 does not determine a prediction direction, a forward direction with respect to the prediction unit 441 And r may be configured to transmit an instruction r for obtaining a predicted value string in both directions. In this case, a prediction direction selection unit 143 is added as indicated by a dotted line in FIG. 15, and the prediction unit 441 obtains a prediction value sequence for both the forward direction and the reverse direction, and thereafter the same as the bidirectional predictive encoding device 100. Perform the process.
With this configuration, the worst calculation amount is the same as that when bidirectional prediction is performed in all cases, but the average calculation amount can be reduced.

The fifth embodiment is configured to enable selection of a prediction direction without transmitting / receiving information directly indicating the prediction direction.
FIG. 16 illustrates a functional configuration example of the bidirectional predictive encoding device 500 of the present invention, and FIG. 17 illustrates a functional configuration example of the bidirectional predictive decoding device 900 provided to face the bidirectional predictive encoding device 500.

  The bi-directional predictive coding apparatus 500 includes a frame dividing unit 110, a prediction coefficient calculating unit 520, a prediction coefficient coding unit 130, a linear prediction unit 540, a prediction error coding unit 150, and a multiplexing unit 160. Further, the prediction coefficient calculation unit 520 includes a prediction analysis unit 521 and a prediction coefficient quantization unit 122, and the linear prediction unit 540 includes a selection rule storage unit 549, a prediction unit 541, and a subtraction unit 142. In other words, except for the absence of the prediction direction selection unit 143, the configuration is the same as that of the bidirectional prediction encoding apparatus 100 except for the prediction analysis unit 521, the selection rule storage unit 549, and the prediction unit 541.

  The selection rule storage unit 549 holds a selection rule for a prediction direction set in advance for a predetermined parameter value used when obtaining a prediction coefficient. Parameters applicable to the present invention may be selected as appropriate. Here, a case where the predetermined parameter is the order P of the prediction coefficient will be described as an example.

  The prediction analysis unit 521 adaptively determines the order P of the prediction coefficient for each frame and outputs it together with the prediction coefficient K. In addition, in this embodiment, the order P of the prediction coefficient is also used as information s indicating the prediction direction selected on the encoding device side, and is necessary for the prediction processing in the decoding device. The information is multiplexed and transmitted together with the information.

Prediction means 541, as a key the order P of the prediction coefficients, to find the selection rule storage unit 549, calculates a predicted value string X _p by the prediction direction s _p that corresponds to the order P.
Subtraction means 142 obtains a prediction error sequence E obtained by subtracting the predicted value sequence X _p from the signal sequence X, prediction error encoding unit 150 generates the encoded coding prediction error column C _e the prediction error sequence E Output.

On the other hand, the bidirectional predictive decoding apparatus 900 includes a demultiplexing unit 610, a prediction coefficient decoding unit 620, a prediction error decoding unit 630, a linear prediction decoding unit 940, and a selection rule storage unit 915. That is, the configuration is the same as that of the bidirectional predictive decoding device 600 except for the selection rule storage unit 915 and the linear predictive decoding unit 940.
Similar to the selection rule storage unit 549 on the encoding side, the selection rule storage unit 915 holds a selection rule of a prediction direction set in advance for a predetermined parameter value used when obtaining a prediction coefficient. Here, as described above, a case where the predetermined parameter is the order P of the prediction coefficient will be described as an example.

The linear predictive decoding unit 940 searches the selection rule storage unit 915 using the value of the information s indicating the prediction direction separated by the demultiplexing unit 610 (in this example, the order P of the prediction coefficient) as a key, and corresponds to the order P. It performs the same processing as the linear prediction decoding unit 640 described in embodiment 1 by the prediction direction s _p to obtain the original signal sequence X.
With the above configuration, even when it is necessary to transmit and receive both information s originally indicating the prediction direction and a predetermined parameter (for example, the order P of the prediction coefficient), information indicating the prediction direction by the predetermined parameter. Since it can also serve as s, encoding efficiency can be improved.

  When each of the above devices is realized by a computer, processing contents of functions that each device should have are described by a program. The processing functions are realized on the computer by executing the program on the computer. In this case, at least a part of the processing content may be realized by hardware. Further, the various processes described above are not only executed in time series according to the description, but may be executed in parallel or individually according to the processing capability of the apparatus that executes the processes or as necessary. In addition, it can change suitably in the range which does not deviate from the meaning of this invention.

  The program describing the processing contents can be recorded on a computer-readable recording medium. The computer-readable recording medium may be any medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, or a semiconductor memory. Specifically, for example, the magnetic recording device may be a hard disk device or a flexible Discs, magnetic tapes, etc. as optical disks, DVD (Digital Versatile Disc), DVD-RAM (Random Access Memory), CD-ROM (Compact Disc Read Only Memory), CD-R (Recordable) / RW (ReWritable), etc. As the magneto-optical recording medium, MO (Magneto-Optical disc) or the like can be used, and as the semiconductor memory, EEP-ROM (Electronically Erasable and Programmable-Read Only Memory) or the like can be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  As an execution form different from the above-described embodiment, the computer may read the program directly from the portable recording medium and execute processing according to the program. Each time is transferred, the processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program in this embodiment includes information that is used for processing by an electronic computer and that conforms to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

Claims (20)

A prediction coefficient calculation unit that obtains a prediction coefficient by performing linear prediction analysis on an input signal composed of a time-series sample sequence in units of frames composed of a plurality of samples;
A prediction coefficient encoding unit that encodes the prediction coefficient and outputs a prediction coefficient code;
The input signal and a prediction coefficient corresponding to the prediction coefficient code are input, and a prediction direction of either a forward direction or a reverse direction is selected in time in the frame, and the input is performed by prediction of the selected prediction direction. A linear prediction unit that outputs a prediction error sequence obtained from a signal and a prediction coefficient corresponding to the prediction coefficient code, and information indicating the selected prediction direction;
A prediction error encoding unit that encodes the prediction error sequence and outputs a prediction error code sequence;
A bi-directional predictive encoding device. The bidirectional predictive encoding device according to claim 1,
The linear prediction unit is a bidirectional predictive coding apparatus that obtains a prediction error sequence in both the forward direction and the reverse direction, and selects a prediction direction with a smaller code amount or a smaller estimated value of the code amount. The bidirectional predictive encoding device according to claim 1,
When the value of the prediction coefficient or the value of the predetermined parameter used when obtaining the prediction coefficient is within a preset range, the linear prediction unit predicts a prediction error sequence for both the forward direction and the reverse direction. Is selected, and the prediction direction with the smaller code amount or the estimated value of the code amount is selected, and otherwise, one of the predetermined directions is selected as the prediction direction. The bidirectional predictive coding apparatus according to any one of claims 1 to 3,
The prediction coefficient calculation unit adaptively determines the prediction order of the prediction coefficient for each frame,
A bidirectional predictive encoding apparatus, further comprising: a variable length encoding unit that integrally performs variable length encoding and outputs the information indicating the prediction order and the selected prediction direction. The bidirectional predictive encoding device according to claim 1,
The prediction coefficient calculation unit calculates the prediction coefficient by a sequential analysis method using forward prediction and backward prediction, and outputs the prediction coefficient together with a forward prediction error and a backward prediction error obtained in the process of the sequential analysis method,
The linear prediction unit selects the forward direction as the prediction direction when the forward prediction error is smaller than the backward prediction error, and selects the reverse direction as the prediction direction when the backward prediction error is smaller than the forward prediction error. A bidirectional predictive encoding device. The bidirectional predictive encoding device according to claim 1,
The information indicating the prediction direction is a value of a predetermined parameter used when obtaining the prediction coefficient,
The linear prediction unit includes selection rule storage means for holding a selection rule for a prediction direction set in advance for the value of the predetermined parameter, and the prediction direction is determined according to the value of the predetermined parameter according to the selection rule. Bidirectional predictive encoding device to select. A prediction coefficient decoding unit that decodes the prediction coefficient from the prediction coefficient code;
A prediction error sequence decoding unit that decodes the prediction error sequence from the prediction error code sequence;
Based on the prediction coefficient, the prediction error sequence, and information indicating the prediction direction of the prediction error sequence, a linear prediction decoding unit that restores the original signal by prediction of the prediction direction;
A bi-directional predictive decoding device. The bidirectional predictive decoding device according to claim 7,
A variable-length code decoding unit that separates the prediction order of the prediction coefficient that is integrally variable-length encoded and information indicating the prediction direction of the prediction error sequence;
The linear predictive decoding unit is a bi-directional predictive decoding device that further receives the prediction order and performs prediction using the prediction order. The bidirectional predictive decoding device according to claim 7,
The information indicating the prediction direction is a value of a predetermined parameter used when obtaining the prediction coefficient,
A selection rule storage unit that holds a prediction direction selection rule set in advance for the value of the predetermined parameter;
The linear predictive decoding unit is a bidirectional predictive decoding apparatus that determines the prediction direction according to a value of the predetermined parameter according to the selection rule. A prediction coefficient calculation step for obtaining a prediction coefficient by performing linear prediction analysis on an input signal composed of a time-series sample sequence in units of frames consisting of a plurality of samples;
A prediction coefficient encoding step of encoding the prediction coefficient and outputting a prediction coefficient code;
From the input signal and a prediction coefficient corresponding to the prediction coefficient code, a prediction direction of either the forward direction or the reverse direction in time is selected in the frame, and the input signal is predicted by the prediction of the selected prediction direction. A linear prediction step for outputting a prediction error sequence obtained from a prediction coefficient corresponding to the prediction coefficient code, and information indicating the selected prediction direction;
A prediction error encoding step of encoding the prediction error sequence and outputting a prediction error code;
A bi-directional predictive coding method for performing The bidirectional predictive encoding method according to claim 10,
The linear prediction step is a bidirectional predictive encoding method in which a prediction error sequence is obtained for both the forward direction and the reverse direction, and the prediction direction with the smaller code amount or the estimated code amount is selected. The bidirectional predictive encoding method according to claim 10,
In the linear prediction step, when the value of the prediction coefficient or the value of a predetermined parameter used when obtaining the prediction coefficient is within a preset range, prediction error sequences for both the forward direction and the reverse direction The prediction direction with the smaller code amount or the estimated value of the code amount is selected, and in other cases, one of the predetermined directions is selected as the prediction direction. The bidirectional predictive encoding method according to any one of claims 10 to 12,
The prediction coefficient calculation step adaptively determines the prediction order of the prediction coefficient for each frame,
A bidirectional predictive coding method further executing a variable length coding step of integrally variable length coding and outputting the information indicating the prediction order and the information indicating the selected prediction direction. The bidirectional predictive encoding method according to claim 10,
The prediction coefficient calculation step calculates the prediction coefficient by a sequential analysis method using forward prediction and backward prediction, and outputs the prediction coefficient together with the forward prediction error and the backward prediction error obtained in the process of the sequential analysis method,
The linear prediction step selects the forward direction as the prediction direction when the forward prediction error is smaller than the backward prediction error, and selects the reverse direction as the prediction direction when the backward prediction error is smaller than the forward prediction error. Bidirectional predictive encoding method. The bidirectional predictive encoding method according to claim 10,
The information indicating the prediction direction is a value of a predetermined parameter used when obtaining the prediction coefficient,
The bidirectional prediction encoding method, wherein the linear prediction step selects the prediction direction according to a value of the predetermined parameter in accordance with a prediction direction selection rule set in advance for the value of the predetermined parameter. A prediction coefficient decoding step of decoding a prediction coefficient from the prediction coefficient code;
A prediction error sequence decoding step of decoding a prediction error sequence from the prediction error code sequence;
Based on the prediction coefficient, the prediction error sequence, and information indicating the prediction direction of the prediction error sequence, a linear prediction decoding step of restoring the original signal by prediction of the prediction direction;
A bi-directional predictive decoding method. The bidirectional predictive decoding method according to claim 16,
A variable length code decoding step for separating the prediction order of the prediction coefficient integrally variable length encoded and the information indicating the prediction direction of the prediction error sequence;
The linear predictive decoding step is a bidirectional predictive decoding method for obtaining a predictive value sequence with the predictive order. The bidirectional predictive decoding method according to claim 16,
The information indicating the prediction direction is a value of a predetermined parameter used when obtaining the prediction coefficient,
The bidirectional predictive decoding method, wherein the linear predictive decoding step determines the prediction direction according to a value of the predetermined parameter in accordance with a selection rule of a prediction direction set in advance for the value of the predetermined parameter.   A program for causing a computer to function as the bidirectional predictive coding apparatus according to any one of claims 1 to 6 or the bidirectional predictive decoding apparatus according to any one of claims 7 to 9.   The computer which recorded the program for functioning a computer as the bidirectional | two-way prediction encoding apparatus in any one of Claims 1 thru | or 6 or the bidirectional | two-way prediction decoding apparatus in any one of Claims 7 thru | or 9 is readable. recoding media.

Images