JP2004015537A - Audio signal encoding device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an audio signal encoding device for removing the cause of sound quality deterioration which occurs when a quantizing noise included in an audio signal is suppressed. <P>SOLUTION: The device comprises: a prediction gain calculating means 110 for generating a linear prediction coefficient of discrete coefficient information in each frequency at every block, and calculating a prediction gain being the ratio of a linear prediction coefficient level and the level of an error in linear prediction; a threshold determining means 120 for determining whether each calculated prediction gain exceeds a prescribed threshold; a block selecting means 130 for selecting the number of blocks which is not greater than a prescribed number from the objective blocks corresponding to the prediction gain which is judged to exceed the threshold; and a quantizing noise suppressing means 140 for filtering each piece of the discrete coefficient information, suppressing the quantizing noise and generating the signal of a prediction residual difference being a linear prediction error at every block. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an audio signal encoding device that encodes an audio signal and suppresses quantization noise included in the audio signal.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, an audio signal encoding apparatus of this type that conforms to an audio standard such as MPEG (Moving Picture Experts Group), which is a standard for moving images and audio, has been widely used.
TNS (Temporal Noise Shaping), which is a process for suppressing quantization noise defined by the MPEG audio standard, divides an audio signal into a plurality of blocks and quantizes the divided blocks according to the block length. A target frequency, a prediction coefficient length, or a prediction gain threshold for noise suppression processing is determined. Further, a prediction gain is calculated by using a known linear prediction analysis based on an MDCT (Modified Discrete Cosine Transform) coefficient which is a kind of discrete cosine coefficient included in the block. The prediction gain is compared with the above-mentioned prediction gain threshold value, and whether or not the prediction gain is larger than the threshold value is compared. After converting the calculated prediction coefficient into a reflection coefficient, a filter coefficient, which is a coefficient obtained by nonlinear quantization, is generated. The generated filter coefficients are subjected to a filtering process on the MDCT coefficients to generate a prediction residual signal. By these processes, the quantization noise is uniformly distributed on the whole time axis, but since the quantization noise is distributed to a place where the output level is large on the time axis by the above filter processing, the pre-echo is reduced. can do.
[0003]
[Problems to be solved by the invention]
However, in such a conventional audio signal encoding device conforming to the MPEG audio standard, when an audio signal is transmitted at a low transmission rate, if the quantization noise included in the audio signal is excessively suppressed, the Since the amount of information generated due to the suppression increases, there is a problem that information for encoding the frequency component decreases and the sound quality deteriorates at the time of decoding. In addition, since the encoding is performed ignoring the properties of the audio signal, there is a problem that the audio quality is deteriorated when the audio signal is decoded.
The present invention has been made in order to solve such a problem, and provides an audio signal encoding device capable of reducing sound quality degradation that occurs when suppressing quantization noise included in an audio signal. Things.
[0004]
[Means for Solving the Problems]
The audio signal encoding apparatus according to the present invention divides an audio signal into a plurality of blocks, and converts each block into blocks based on discrete coefficient information that is information of a frequency domain obtained by converting each block. Predictive gain calculating means for performing linear prediction on the discrete coefficient information to generate a linear prediction coefficient, and calculating a prediction gain, which is a ratio between the level of the linear prediction coefficient and the level of the error of the linear prediction, for each of the blocks And a threshold determination unit that determines whether each of the prediction gains calculated for each of the blocks exceeds a predetermined threshold, and a target block that is a block corresponding to the prediction gain that is determined to have exceeded the threshold. From among them, block selecting means for selecting a number of blocks equal to or less than a predetermined number, based on a filter coefficient obtained from the selected target block. And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is an error of linear prediction for each block, wherein an order of the linear prediction coefficient is , A predetermined order or less. With this configuration, a target block that is a block whose prediction gain has been determined to exceed the threshold value is selected, and quantization noise included in the audio signal is suppressed because quantization noise included in the audio signal is not excessively suppressed. It is possible to reduce the sound quality deterioration that occurs at the time.
[0005]
Further, the audio signal encoding device of the present invention divides the audio signal into a plurality of blocks, and converts A prediction gain that performs linear prediction on discrete coefficient information at a frequency to generate a linear prediction coefficient, and calculates a prediction gain that is a ratio between a level of the linear prediction coefficient and an error level of the linear prediction for each of the blocks; Calculating means, threshold value determining means for determining whether each of the predicted gains calculated for each of the blocks has exceeded a predetermined threshold, and an object which is a block corresponding to the predicted gain determined to have exceeded the threshold. A block selecting means for selecting a number of blocks equal to or less than a predetermined number from the blocks; And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is a linear prediction error for each block. The order is equal to or less than a maximum order which is a predetermined order, and the prediction gain calculating means generates statistical information based on the discrete coefficient information, determines a frequency band range according to the statistical information, and determines The prediction gain is calculated for each of the blocks based on the discrete coefficient information for each of the blocks included in the range of the set frequency band. With this configuration, statistical information is generated based on the discrete coefficient information, and the frequency component is encoded according to the properties of the audio signal. Therefore, it is possible to suppress sound quality degradation during decoding.
[0006]
Further, the audio signal encoding apparatus of the present invention divides an audio signal into a plurality of blocks, and converts each of the blocks into blocks based on each discrete coefficient information that is frequency domain information obtained by converting the blocks. A prediction gain that performs linear prediction on discrete coefficient information at a frequency to generate a linear prediction coefficient, and calculates a prediction gain that is a ratio between a level of the linear prediction coefficient and an error level of the linear prediction for each of the blocks; Calculating means, threshold value determining means for determining whether each of the predicted gains calculated for each of the blocks has exceeded a predetermined threshold, and an object which is a block corresponding to the predicted gain determined to have exceeded the threshold. A block selecting means for selecting a number of blocks equal to or less than a predetermined number from the blocks; And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is a linear prediction error for each block. The order is equal to or less than the maximum order, which is a predetermined order, and further includes a compression ratio calculating unit that calculates a compression ratio based on a sampling frequency, a transmission speed, and the number of channels, and the block selecting unit is configured to change the compression ratio according to the compression ratio The number of selections that is the number for selecting the target block is determined, and the number of blocks equal to or less than the determined number of selections is selected. With this configuration, it is possible to automatically determine the number of selections in accordance with the compression ratio when transmitting information obtained by encoding the audio signal.
[0007]
Further, the audio signal encoding device of the present invention divides the audio signal into a plurality of blocks, and converts A prediction gain that performs linear prediction on discrete coefficient information at a frequency to generate a linear prediction coefficient, and calculates a prediction gain that is a ratio between a level of the linear prediction coefficient and an error level of the linear prediction for each of the blocks; Calculating means, threshold value determining means for determining whether each of the predicted gains calculated for each of the blocks has exceeded a predetermined threshold, and an object which is a block corresponding to the predicted gain determined to have exceeded the threshold. A block selecting means for selecting a number of blocks equal to or less than a predetermined number from the blocks; And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is a linear prediction error for each block. The order is equal to or less than a maximum order which is a predetermined order, and the prediction gain calculating means determines the maximum order according to the compression ratio, and sets the prediction gain for each of the blocks based on the determined maximum order. It has a configuration for calculating. With this configuration, it is possible to automatically determine the maximum order in accordance with the compression ratio when transmitting information obtained by encoding an audio signal.
[0008]
Further, the audio signal encoding apparatus of the present invention divides an audio signal into a plurality of blocks, and converts each of the blocks into blocks based on each discrete coefficient information that is frequency domain information obtained by converting the blocks. A prediction gain that performs linear prediction on discrete coefficient information at a frequency to generate a linear prediction coefficient, and calculates a prediction gain that is a ratio between a level of the linear prediction coefficient and an error level of the linear prediction for each of the blocks; Calculating means, threshold value determining means for determining whether each of the predicted gains calculated for each of the blocks has exceeded a predetermined threshold, and an object which is a block corresponding to the predicted gain determined to have exceeded the threshold. A block selecting means for selecting a number of blocks equal to or less than a predetermined number from the blocks; And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is a linear prediction error for each block. The degree is equal to or less than the maximum degree which is a predetermined degree, and the threshold value determining means determines a threshold value according to the compression ratio and determines whether or not the determined threshold value is exceeded. With this configuration, it is possible to automatically determine the threshold value according to the compression ratio at the time of transmitting the information obtained by encoding the audio signal.
[0009]
Further, the audio signal encoding apparatus of the present invention divides an audio signal into a plurality of blocks, and converts each of the blocks into blocks based on each discrete coefficient information that is frequency domain information obtained by converting the blocks. A prediction gain that performs linear prediction on discrete coefficient information at a frequency to generate a linear prediction coefficient, and calculates a prediction gain that is a ratio between a level of the linear prediction coefficient and an error level of the linear prediction for each of the blocks; Calculating means, threshold value determining means for determining whether each of the predicted gains calculated for each of the blocks has exceeded a predetermined threshold, and an object which is a block corresponding to the predicted gain determined to have exceeded the threshold. A block selecting means for selecting a number of blocks equal to or less than a predetermined number from the blocks; And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is a linear prediction error for each block. The order is equal to or less than a maximum order which is a predetermined order, and the threshold value judging means determines a threshold value when the compression ratio is equal to or less than a predetermined value and according to the order of the linear prediction coefficient. Has been determined to determine whether the threshold value has been exceeded. With this configuration, it is possible to automatically determine the threshold value according to the compression ratio when transmitting information obtained by encoding the audio signal and the order of the linear prediction coefficient.
[0010]
Further, the audio signal encoding apparatus of the present invention divides an audio signal into a plurality of blocks, and converts each of the blocks into blocks based on each discrete coefficient information that is frequency domain information obtained by converting the blocks. A prediction gain that performs linear prediction on discrete coefficient information at a frequency to generate a linear prediction coefficient, and calculates a prediction gain that is a ratio between a level of the linear prediction coefficient and an error level of the linear prediction for each of the blocks; Calculating means, threshold value determining means for determining whether each of the predicted gains calculated for each of the blocks has exceeded a predetermined threshold, and an object which is a block corresponding to the predicted gain determined to have exceeded the threshold. A block selecting means for selecting a number of blocks equal to or less than a predetermined number from the blocks; And a quantization noise suppressing unit that suppresses quantization noise by filtering the discrete coefficient information and generates a prediction residual signal that is a linear prediction error for each block. The order is equal to or less than a maximum order which is a predetermined order, and further, the prediction residual signal generated by the quantization noise suppressing means is restored according to the compression ratio, and the restored information and a base of the restoration are restored. And a means for outputting a signal of the prediction residual or each of the discrete coefficient information according to the result of the determination. With this configuration, an error between discrete coefficient information and information obtained by restoring a signal of a prediction residual in which quantization noise has been suppressed is determined, and information to be output is determined, so that sound quality degradation is reduced during decoding. It is possible to do.
[0011]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing a block configuration of an audio signal encoding device according to a first embodiment of the present invention. The audio signal encoding apparatus 100 according to the first embodiment of the present invention includes a prediction gain calculation unit 110 that generates a prediction gain, a threshold determination unit 120 that determines whether the prediction gain exceeds a predetermined threshold, a prediction gain Is comprised of a block selecting means 130 for selecting a predetermined number of blocks exceeding a predetermined threshold, and a quantization noise suppressing means 140 for suppressing quantization noise included in the block.
[0012]
The prediction gain calculation unit 110 is a unit that inputs each piece of discrete coefficient information that is information in the frequency domain obtained by conversion for each block, and generates the following linear prediction coefficient and prediction gain. Here, a block is obtained by dividing an audio signal into a plurality. The discrete coefficient information may be an MDCT (Modified Discrete Cosine Transform) coefficient described in an MPEG standard or the like.
[0013]
The prediction gain calculation means 110 performs linear prediction of the discrete coefficient information at each frequency for each block based on the input discrete coefficient information to generate a linear prediction coefficient. The method of generating the linear prediction coefficients may be a known Levinson-Durbin algorithm or another method. In addition, the order of the linear prediction is generated so as to be equal to or less than the maximum order which is a predetermined order.
[0014]
The prediction gain calculation unit 110 calculates a prediction gain, which is a ratio between the level of the generated linear prediction coefficient and the level of the error of the linear prediction, for each block, and outputs the prediction gain corresponding to each block to the threshold determination unit 120. . The prediction gain indicates the possibility that the block contains quantization noise. If the probability is high, the prediction gain increases. The method of calculating the prediction gain is described in an MPEG standard or the like.
[0015]
Further, the prediction gain calculating means 110 generates statistical information based on each discrete coefficient information, determines a frequency band range according to the statistical information, and determines each block included in the determined frequency band range. May be calculated for each of the blocks based on the discrete coefficient information. As a method of determining the range of the frequency band, the range of the frequency band may be narrowed according to the magnitude of dispersion of the spectrum included in the scale factor band, which is a unit for dividing the frequency axis.
[0016]
The threshold determination unit 120 receives the prediction gain corresponding to each block output by the prediction gain calculation unit 110 as input, determines whether each of the input prediction gains exceeds a predetermined threshold, and outputs information on the determination result. (Hereinafter, referred to as judgment information) to the block selecting means 130. The determination information may be information indicating whether or not a predetermined threshold value given to each block is exceeded.
[0017]
The block selecting unit 130 receives the judgment information output by the threshold judging unit 120 as an input and, based on the judging information, selects a prediction block obtained from a block among target blocks whose prediction gains are judged to exceed the threshold. This is a means for selecting in order from the one having the largest gain so as to be equal to or less than the predetermined number which is the selected number. For example, if the number of target blocks determined to exceed the threshold is 5 and the number of selections is 3, three blocks are selected in order from the one with the largest prediction gain. If the number of target blocks determined to exceed the threshold is 2 and the number of selections is 3, two blocks are selected.
[0018]
The block selection unit 130 outputs information on the result of the selection (hereinafter, referred to as selection information) and each block to the quantization noise suppression unit 140. The selection information may be information indicating whether or not a selection is given to each block.
[0019]
The quantization noise suppressing unit 140 receives the selection information output by the block selecting unit 130 as input, calculates a filter coefficient from the selected block, and filters the discrete coefficient information with a filter to which the calculated filter coefficient is applied. , Means for suppressing quantization noise.
[0020]
The quantization noise suppressing unit 140 generates and outputs a prediction residual signal which is an error of the linear prediction for each block. TNS (Temporal Noise Shaping), which is a process for suppressing quantization noise, is described in MPEG standards and the like.
[0021]
As described above, the audio signal encoding device according to the first embodiment of the present invention selects a target block that is a block whose prediction gain has been determined to have exceeded the threshold, and selects a target block included in the audio signal. Since the quantization noise is not excessively suppressed, it is possible to reduce the sound quality deterioration that occurs when the quantization noise included in the audio signal is suppressed. In addition, since statistical information is generated based on the discrete coefficient information and the audio signal is encoded according to the properties of the audio signal, it is possible to reduce sound quality deterioration during decoding.
[0022]
FIG. 2 is a diagram illustrating a block configuration of an audio signal encoding device according to a second embodiment of the present invention. The audio signal encoding device 200 according to the second embodiment of the present invention includes a prediction gain calculation unit 210 that generates a prediction gain, a threshold determination unit 220 that determines whether the prediction gain exceeds a predetermined threshold, and a block. A block selection unit 230 for selecting a predetermined number, a quantization noise suppression unit 140 for suppressing quantization noise included in a block, and a compression ratio calculation unit 250 for calculating a compression ratio based on a sampling frequency, a transmission speed, and the number of channels. It is constituted by.
Note that, of the units configuring the audio signal encoding device 200 according to the second embodiment of the present invention, the same units as the units configuring the audio signal encoding device 100 according to the first embodiment of the present invention. Those performing the processing are denoted by the same reference numerals, and description thereof will be omitted.
[0023]
The prediction gain calculation unit 210 receives the compression ratio output from the compression ratio calculation unit 250 and discrete coefficient information that is information in the frequency domain obtained by converting each block, and calculates the following linear prediction coefficient, And a means for generating a gain. Here, a block is obtained by dividing an audio signal into a plurality. The discrete coefficient information may be an MDCT (Modified Discrete Cosine Transform) coefficient described in an MPEG standard or the like.
[0024]
Further, the prediction gain calculation means 210 determines the maximum order according to the input compression ratio. For example, as a method of determining the maximum order, a table in which the compression rate and the maximum order correspond may be created in advance, and the maximum order may be obtained according to this table.
[0025]
The prediction gain calculation means 210 performs linear prediction of the discrete coefficient information at each frequency for each block based on the input discrete coefficient information to generate a linear prediction coefficient. The method of generating the linear prediction coefficients may be a known Levinson-Durbin algorithm or another method. The order of the linear prediction coefficient is generated so as to be equal to or less than the determined maximum order.
[0026]
The prediction gain calculation unit 210 calculates a prediction gain, which is a ratio between the level of the generated linear prediction coefficient and the level of the error of the linear prediction, for each block, and outputs the prediction gain corresponding to each block to the threshold determination unit 220. . The prediction gain indicates the possibility that the block contains quantization noise. If the probability is high, the prediction gain increases. The method of calculating the prediction gain is described in an MPEG standard or the like.
[0027]
Further, the prediction gain calculating means 210 generates statistical information based on each discrete coefficient information, determines a frequency band range according to the statistical information, and calculates a frequency band range for each of the blocks included in the determined frequency band range. May be calculated for each of the blocks based on the discrete coefficient information. As a method of determining the range of the frequency band, the range of the frequency band may be narrowed according to the magnitude of dispersion of the spectrum included in the scale factor band, which is a unit for dividing the frequency axis.
[0028]
The threshold value determining unit 220 receives the compression ratio output from the compression ratio calculating unit 250 and the prediction gain corresponding to each block output from the prediction gain calculating unit 110, and determines a threshold value according to the input compression ratio. It is a means to do. For example, as a method of determining the threshold, a table in which the compression ratio and the threshold correspond to each other may be created in advance, and the threshold may be determined according to this table.
[0029]
Further, when the input compression ratio is equal to or less than a predetermined value, the threshold determination unit 220 determines the threshold according to the order of the linear prediction coefficient for each block. For example, when the order of the linear prediction coefficient is large, when suppressing the quantization noise, the amount of information generated due to the suppression may increase. Therefore, when the order of the linear prediction coefficient is large, the threshold value is determined to be a large value.
[0030]
The threshold value determining means 220 determines whether or not each of the input prediction gains exceeds the threshold value determined as described above, and determines information relating to the result of the determination (hereinafter, referred to as determination information) and each block with the block selecting means 230. Output to The determination information may be information indicating whether or not a predetermined threshold value given to each block is exceeded.
[0031]
The block selection unit 230 receives the compression ratio output by the compression ratio calculation unit 250 and the determination information output by the threshold determination unit 220 as input, and determines a predetermined number of selections according to the input compression ratio. It is a means to do. For example, as a method of determining the number of selections, a table in which the compression ratio and the number of selections correspond may be created in advance, and the number of selections may be obtained according to this table.
[0032]
In addition, based on the input determination information, the block selection unit 230 determines a predetermined order from the target block whose prediction gain has been determined to have exceeded the threshold value, in descending order of prediction gain obtained from the block. The number is selected so as to be equal to or less than the selected number.
[0033]
The block selection unit 230 outputs information on the result of the selection (hereinafter, referred to as selection information) and each block to the quantization noise suppression unit 140. The selection information may be information indicating whether or not a selection is given to each block.
[0034]
The compression ratio calculation unit 250 is a unit that calculates a compression ratio based on the sampling frequency, the transmission speed, and the number of channels, and outputs the calculated compression ratio to a unit that requires the compression ratio. For example, when the sampling frequency is represented by fs, the transmission speed is represented by br, and the number of channels is represented by ch, the compression ratio is obtained by the following equation.
Compression rate = (fs / br) / ch
[0035]
As described above, the audio signal encoding apparatus according to the second embodiment of the present invention automatically sets the maximum order, the threshold, or the degree according to the compression rate when transmitting information obtained by encoding the audio signal. The number of selections can be determined.
[0036]
FIG. 3 is a diagram illustrating a block configuration of an audio signal encoding device according to a third embodiment of the present invention. The audio signal encoding apparatus 300 according to the third embodiment of the present invention includes a prediction gain calculation unit 210 that generates a prediction gain, a threshold determination unit 220 that determines whether the prediction gain exceeds a predetermined threshold, and a block. A block selection unit 230 for selecting a predetermined number, a quantization noise suppression unit 140 for suppressing quantization noise included in a block, a compression ratio calculation unit 250 for calculating a compression ratio, and a prediction residual signal with quantization noise suppressed The decoding decision means 360 decides whether or not to output the signal of the prediction residual based on
Note that, of the units configuring the audio signal encoding device 300 according to the third embodiment of the present invention, the same units as the units configuring the audio signal encoding device 200 according to the second embodiment of the present invention. Those performing the processing are denoted by the same reference numerals, and description thereof will be omitted.
[0037]
The decoding determining unit 360 receives the input of the prediction residual signal output by the quantization noise suppressing unit 140 and the compression ratio output by the compression ratio calculating unit 250 output by the prediction gain calculating unit 210. Means for processing such information.
[0038]
Further, the decoding determining means 360 calculates the information obtained by restoring the signal of the prediction residual, and the information for each block which is the information of the frequency domain obtained by converting the audio signal included in the block for each block. An error with the discrete coefficient information is calculated.
[0039]
For example, as a method of restoring the signal of the prediction residual, only the number of spectra included in the signal of the prediction residual is restored according to the compression ratio.In other words, when the compression ratio is very small, the number of spectra that can be quantized is small. Therefore, when the compression ratio is small, the number of spectra to be restored is also reduced. In addition, at the time of encoding, in the scale factor band, which is a unit for dividing the frequency axis, the absolute value of the spectrum is selected in descending order, and the signal of the prediction residual is restored.
For example, as a method of calculating an error between the information obtained by restoring the signal of the prediction residual and the discrete coefficient information, the calculation is performed using an energy value or the like included in each scale factor band.
[0040]
The decoding determining means 360 determines the calculated error and outputs a signal of a prediction residual or discrete coefficient information. For example, if the error between the information obtained by restoring the signal of the prediction residual obtained by the energy value included in each of the scale factor bands and the discrete coefficient information exceeds a reference value, when decoding, In order to cause sound quality deterioration, discrete coefficient information is output without outputting a prediction residual signal. If the error is within the reference value, a signal of the prediction residual is output.
[0041]
As described above, the audio signal encoding apparatus according to the third embodiment of the present invention calculates the error between the discrete coefficient information and the information obtained by restoring the signal of the prediction residual in which quantization noise is suppressed. Since the information to be determined and output is determined, it is possible to suppress sound quality deterioration during decoding.
[0042]
【The invention's effect】
As described above, the present invention can provide an audio signal encoding apparatus that reduces sound quality degradation that occurs when suppressing quantization noise included in an audio signal.
[Brief description of the drawings]
FIG. 1 is a diagram showing a block configuration of an audio signal encoding device according to a first embodiment of the present invention.
FIG. 2 is a diagram showing a block configuration of an audio signal encoding device according to a second embodiment of the present invention.
FIG. 3 is a diagram showing a block configuration of an audio signal encoding device according to a third embodiment of the present invention.
[Explanation of symbols]
100, 200, 300 audio signal encoding device
110, 210 Predicted gain calculating means
120, 220 threshold determination means
130, 230 block selecting means
140 Quantization noise suppression means
250 Compression ratio calculation means
360 decryption determination means

Claims (7)

The audio signal is divided into a plurality of blocks, and based on each discrete coefficient information which is frequency domain information obtained by converting each block, linear prediction is performed on the discrete coefficient information at each frequency for each block. Predictive gain calculating means for calculating, for each of the blocks, a predictive gain that is a ratio between the level of the linear predictive coefficient and the level of the error of the linear prediction; and Threshold value determining means for determining whether or not each prediction gain has exceeded a predetermined threshold; and a target number of blocks equal to or less than a predetermined number among target blocks corresponding to the prediction gains determined to have exceeded the threshold. And selecting each discrete coefficient information based on a filter coefficient obtained from the selected target block. Quantization noise suppressing means for suppressing the quantization noise and generating a prediction residual signal which is an error of the linear prediction for each block, wherein an order of the linear prediction coefficient is a maximum order which is a predetermined order. An audio signal encoding device characterized by the following. The predictive gain calculating means generates statistical information based on the discrete coefficient information, determines a frequency band range according to the statistical information, and calculates a discrete value for each block included in the determined frequency band range. The audio signal encoding device according to claim 1, wherein the prediction gain is calculated for each of the blocks based on coefficient information. A compression ratio calculation unit that calculates a compression ratio based on a sampling frequency, a transmission speed, and the number of channels is provided, and the block selection unit determines a selection number that is a number for selecting the target block according to the compression ratio. 3. The audio signal encoding apparatus according to claim 1, wherein a number of blocks equal to or less than the determined number of selections are selected. 4. The audio signal according to claim 3, wherein the prediction gain calculation unit determines the maximum order according to the compression ratio, and calculates the prediction gain for each of the blocks based on the determined maximum order. Encoding device. The audio signal encoding device according to claim 3, wherein the threshold value determining unit determines a threshold value according to the compression ratio, and determines whether the threshold value is exceeded. The threshold value determining means, when the compression ratio is equal to or less than a predetermined value, and determines a threshold value according to the order of the linear prediction coefficient, and determines whether or not exceeds the determined threshold value. The audio signal encoding device according to claim 3 or 4, wherein: According to the compression ratio, the signal of the prediction residual generated by the quantization noise suppression unit is restored, and an error between the restored information and each of the discrete coefficient information on which the restoration is based is determined. 7. The audio signal encoding apparatus according to claim 3, further comprising a unit that outputs the signal of the prediction residual or the discrete coefficient information according to a result of the determination. .

Images