JPH09261064A - Encoder and decoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain coded and decoded signals without the power loss of an original signal, by providing a bit assigning means and a quantization means, etc., quantizing plural partial band signals generated by a band dividing means based on bit assigning information generated by this assigning means. SOLUTION: A band dividing part 1 divides an input signal into plural partial bands and outputs a partial band signal. An acoustic sense model part 3 spectrum-analyzes the input signal, furthermore analyzes it based on human acoustic sense characteristic and calculates an evaluation function for executing optimum bit-assigning by a bit assigning part 4 in consideration of a maximum value from a maximum value detection part 2. The part 4 decides optimum bit assignment for each partial band based on this evaluation function. A quantization part 5 normalizes the partial band signal in each partial band from the part 1 by the maximum value from the part 2 for improving quantization efficiency and quantizes and encodes it according to bit assignment from the part 4.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoder and a decoder for encoding an input signal (for example, an audio signal such as voice or musical sound), transmitting or recording / reproducing it, and decoding it to obtain a reproduced signal. Is.

[0002]

2. Description of the Related Art There are various methods for high-efficiency coding of audio signals and the like, and one of them is a band division coding method.
Normally, a signal on the time axis is divided into a plurality of partial bands by the filter and encoded while being on the time axis, which is called a band division coding method. The so-called orthogonal transform coding method, in which the data is converted (orthogonally changed) into a plurality of partial bands and coded, is also a kind of band division coding method.

As a filter for the above band division,
For example, a digital filter such as a polyphase filter bank or a QMF filter can be used. Further, as the orthogonal transform, for example, fast Fourier transform (FFT), discrete cosine transform (DCT) and the like can be mentioned.

Further, when the sample data for each partial band divided into a plurality of partial bands is encoded, a predetermined bit allocation is made for each partial band or an adaptive bit allocation is made for each partial band. Encoding is performed.

As such an encoding method, for example, JP-A-04-177300 and JP-A-05-37396 are available.
There is one shown in Japanese Patent Publication No. FIG. 27 is a diagram showing the configuration of the conventional encoder as described above. 1 is a band division unit, 2 is a maximum value detection unit, 3 is an auditory model unit, 4 is a bit allocation unit, 5 is a quantization unit, and 6 is a multiplexing unit.

The operation of the conventional encoder shown in FIG. 27 will be described. The band division unit 1 divides the input signal into a plurality of partial bands and outputs a partial band signal for a certain specific time segment. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band.

The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. When considering the difference between the maximum value of the signal level and the minimum value of the mask characteristic as the evaluation function, bits are sequentially allocated from the partial band having the large difference. When such an evaluation function is used, if this value is negative, control may be added to set the bit allocation to zero in consideration of coding efficiency. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed.

FIG. 28 is a simplified representation of the power distribution of the partial band signal. FIG. 29 shows how bits are allocated by the bit allocation unit 4. The information of each subband signal shown in FIG. 28 is not transmitted when the bit allocation is zero. Therefore, the partial band signal transmitted when the bit allocation shown in FIG. 29 is allocated is as shown in FIG.

FIG. 31 is a diagram showing the structure of a conventional decoder. Reference numeral 1 is a separation unit, 2 is an inverse quantization unit, and 3 is a band synthesis unit. The input encoded data is separated by the separating unit 1 into sample information, maximum value information, and bit allocation information. The inverse quantizer 2 decodes the partial band signal for each partial band whose bit allocation is not zero from each information, substitutes the zero signal as the partial band signal for the partial band whose bit allocation is zero, and The partial band signal is band-combined by the band combining unit 3 into a signal having the original bandwidth.

[0010]

In such a conventional encoder and decoder, the original signal is divided into a plurality of partial bands, and each partial band signal information is transmitted and processed. If the assigned bit is zero, then
The partial band signal information, for example, the maximum value information and the sample information described above, will not be transmitted, so that there is a problem that the decoded signal synthesized after transmission has a reduced power with respect to the original signal.

The present invention has been made to solve the above problems, and guarantees the transmission of necessary partial band signal information, or compensates the power of the partial band signal loss in advance, or the partial band. An object of the present invention is to provide a coding / decoding device that obtains a coded / decoded signal without signal power loss from an original signal by later compensating for the power of the signal loss.

[0012]

An encoder according to claim 1 of the present invention is a band dividing means for dividing an input signal into a plurality of partial bands to generate a plurality of partial band signals, and the input signal to a human being. The spectrum analysis based on the masking rule of the auditory characteristic of, the auditory model means for calculating the evaluation function for the plurality of partial bands, and for the partial bands,
A reference bit allocation table storing predetermined reference bit allocation information, a plurality of parts generated by the band dividing means based on the reference bit allocation information of the reference bit allocation table and the evaluation function calculated by the auditory model means. Bit allocation means for generating bit allocation information for quantizing the band signal, and quantizing the plurality of partial band signals generated by the band dividing means based on the bit allocation information generated by the bit allocation means And a quantizing means.

An encoder according to claim 2 of the present invention is a band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a plurality of partial parts generated by the band dividing means. The power calculation means for calculating the power of the band signal, the auditory model means for spectrum-analyzing the input signal based on the masking rule of human auditory characteristics and calculating the evaluation function for the plurality of partial bands, and the power calculation means A reference bit allocation table for outputting predetermined reference bit allocation information for the partial band based on the calculated power of the partial band, and a bit allocation for quantizing the plurality of partial band signals generated by the band dividing means. The information is based on the reference bit allocation information of the reference bit allocation table and the evaluation function calculated by the auditory model means. And bit allocation means for generating come,
Quantizing means for quantizing the plurality of partial band signals generated by the band dividing means based on the bit allocating information determined by the bit allocating means.

An encoder according to claim 3 of the present invention comprises a band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a masking rule of the input signal for human auditory characteristics. Spectrum analysis based on
The auditory model means for calculating the evaluation function for the plurality of partial bands, the reference bit allocation table for outputting predetermined reference bit allocation information for the partial bands based on the spectrum analysis result of the auditory model means, and the band dividing means. Bit allocation means for generating bit allocation for quantizing a plurality of generated sub-band signals based on the reference bit allocation information of the reference bit allocation table and the evaluation function calculated by the auditory model means; and the bit allocation And quantizing means for quantizing the plurality of partial band signals generated by the band dividing means based on the bit allocation information determined by the means.

An encoder according to claim 4 of the present invention is a band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a plurality of partial parts generated by the band dividing means. Maximum value detecting means for detecting the maximum value of the band signal, auditory model means for spectrum-analyzing the input signal based on the masking rule of human auditory characteristics, and calculating the evaluation function for the plurality of partial bands, and the maximum A reference bit allocation table for storing predetermined reference bit allocation information for the partial band based on the maximum value detected by the value detecting means, and a plurality of partial band signals generated by the band dividing means for quantizing A bit allocation is generated based on the reference bit allocation information of the reference bit allocation table and the evaluation function calculated by the auditory model means. And Tsu bit assignments means, based on the bit allocation information determined by the bit allocation means,
And a quantizing means for quantizing the plurality of partial band signals generated by the band dividing means.

An encoder according to claim 5 of the present invention is a band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and the input signal is a masking rule for human auditory characteristics. Spectrum analysis based on
Auditory model means for calculating an evaluation function for the plurality of partial bands, a reference bit allocation table for storing predetermined reference bit allocation information for the partial bands, and a plurality of partial band signals generated by the band dividing means are quantized. Based on the bit allocation information determined by the bit allocation means, and the bit allocation means for generating the bit allocation for conversion based on the reference bit allocation information of the reference bit allocation table and the evaluation function calculated by the auditory model means. A quantizing means for quantizing the plurality of partial band signals generated by the band dividing means,
Local decoding means for decoding the coded data quantized by the quantizing means, first power analyzing means for calculating the power and spectrum of the input signal, and power of the decoded signal decoded by the local decoding means. And a second power analysis means for calculating a spectrum, and in the reference bit allocation means, the bit allocation to the partial band is reduced from the result of the first power analysis means and the second power analysis means. The predetermined reference bit allocation information is output in consideration of the power loss in this case.

An encoder according to claim 6 of the present invention is a band dividing means for dividing an input signal into a plurality of partial bands to generate a plurality of partial band signals, and a masking rule for the human auditory characteristics of the input signal. And a bit allocation for generating bit allocation information for quantizing the plurality of partial band signals generated by the band dividing means. Means for quantizing the plurality of partial band signals generated by the band dividing means based on the bit allocation information determined by the bit allocating means, and the power of the partial band signals of the band dividing means. The power calculating means to be obtained and the adjustment for adjusting the gain of the input signal corresponding to the power of the partial band obtained by the power calculating means. It is obtained by a means.

An encoder according to claim 7 of the present invention divides an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a plurality of band dividing devices. Maximum value detecting means for detecting the maximum absolute value of the sub-band signal, and auditory model means for spectrum-analyzing the input signal based on the masking rule of human auditory characteristics, and calculating an evaluation function for the plurality of sub-bands. A bit allocation unit for generating bit allocation information for quantizing the plurality of partial band signals generated by the band division unit; and by the band division unit based on the bit allocation information determined by the bit allocation unit, Quantizing means for quantizing the generated plurality of partial band signals, and power for obtaining the power of the partial band signals of the band dividing means Means out, in response to sub-bands of the power obtained by the power calculation unit is obtained by an adjusting means for adjusting the gain of the detected maximum value in the maximum value detecting means.

The encoder according to claim 8 of the present invention determines that the bit allocation to the partial band is zero based on the bit allocation information determined by the bit allocation means and the power information of the partial band calculated by the power calculation means. In consideration of the power loss when the maximum value is detected, the gain of the largest maximum value among the maximum values detected by the maximum value detecting means is adjusted.

The encoder according to claim 9 of the present invention uses the power when the bit allocation to the partial band becomes zero from the bit allocation information determined by the bit allocation means and the spectrum analysis result in the auditory model means. It considers loss.

The encoder according to claim 10 of the present invention is
The power loss when the bit allocation to the partial band becomes zero is considered from the bit allocation determined by the bit allocation means and the maximum value for the partial band detected by the maximum value detecting means.

An encoder according to claim 11 of the present invention is
A local decoding means for decoding the coded data, a first power analysis means for calculating the power and spectrum of the input signal, and a second power calculating means for calculating the power and spectrum of the decoded signal decoded by the local decoding means. And a power analysis means, wherein the power calculation means calculates the power loss when the bit allocation information for the partial band becomes zero from the results of the first power analysis means and the second power analysis means. The gain adjusting means adjusts the gain for the input signal according to the power loss calculated by the power calculating means.

A decoder according to claim 12 of the present invention is
Separation means for separating power information, bit allocation information, maximum value information and sample information included in input coded data, and dequantization for decoding a partial band signal from the bit allocation information, maximum value information and sample information Means and
A band synthesizing unit for band-synthesizing the partial band signals decoded by the inverse quantizing unit, and a gain for adjusting the gain of the decoded signal band-synthesized by the band synthesizing unit according to the power information separated by the separating unit. And adjusting means.

A decoder according to claim 13 of the present invention is
Separation means for separating power information, bit allocation information, maximum value information and sample information included in input coded data, and dequantization for decoding a partial band signal from the bit allocation information, maximum value information and sample information Means and
A power loss calculating means for calculating a power loss for a partial band from a band synthesizing means for synthesizing the partial band signals decoded by the dequantizing means, and power information, bit allocation information and maximum value information from the separating section. And a maximum value adjusting means for adjusting the gain of the maximum value information corresponding to the partial band separated by the separating section according to the power loss calculated by the power loss calculating means.

A decoder according to claim 14 of the present invention is
Bit allocation information included in the input encoded data,
Separating means for separating maximum value information and sample information; dequantizing means for decoding a partial band signal from the bit allocation information, maximum value information and sample information; and a subband signal decoded by the dequantizing means. Band synthesizing means for band synthesizing, partial band signal generating means for determining a partial band having zero bit allocation from the bit allocation information separated by the separating means, and generating the partial band signal, and the inverse quantizing means. Calculating means for adding the partial band signal decoded by the above and the partial band signal generated by the partial band signal generating means, and the partial band signal added by the calculating means is combined with the signal of the original bandwidth. Band combining means, wherein the partial band signal generating means generates a partial band signal generated for a partial band having zero bit allocation. As the level, and it generates a noise according to the minimum audible level.

A decoder according to claim 15 of the present invention is
Bit allocation information included in the input encoded data,
Separating means for separating maximum value information and sample information; dequantizing means for decoding a partial band signal from the bit allocation information, maximum value information and sample information; and a subband signal decoded by the dequantizing means. Band synthesizing means for band synthesizing, partial band signal generating means for determining a partial band having zero bit allocation from the bit allocation information separated by the separating means, and generating the partial band signal, and the inverse quantizing means. Calculating means for adding the partial band signal decoded by the above and the partial band signal generated by the partial band signal generating means, and the partial band signal added by the calculating means is combined with the signal of the original bandwidth. Band combining means, wherein the partial band signal generating means generates a partial band signal generated for a partial band having zero bit allocation. As the level, and generates a noise bit allocation from the maximum value information separated has applied a masking threshold level by the sub-band signals not zero by said separating means.

A decoder according to claim 16 of the present invention is
Bit allocation information included in the input encoded data,
Separation means for separating maximum value information and sample information, dequantization means for decoding a partial band signal from the bit allocation information, maximum value information and sample information, and level information of partial band signals of past frames are stored. Level information storing means, band synthesizing means for synthesizing the partial band signals decoded by the inverse quantizing means, and partial band having zero bit allocation from the bit allocation information separated by the separating means, Partial band signal generation means for generating the partial band signal, operation means for adding the partial band signal decoded by the inverse quantization means, and partial band signal generated by the partial band signal generation means, and Band combining means for combining the partial band signals added by the calculating means into a signal of the original bandwidth, and generating the partial band signal In stage, as the level of the sub-band signals generated for the partial band bit allocation is zero, and generates a noise to which the level stored in the level information storing means.

A decoder according to claim 17 of the present invention is
Bit allocation information included in the input encoded data,
Separating means for separating maximum value information and sample information; dequantizing means for decoding a partial band signal from the bit allocation information, maximum value information and sample information; and a subband signal decoded by the dequantizing means. Band synthesizing means for band synthesizing, partial band signal generating means for determining a partial band having zero bit allocation from the bit allocation information separated by the separating means, and generating the partial band signal, and the inverse quantizing means. Calculating means for adding the partial band signal decoded by the above and the partial band signal generated by the partial band signal generating means, and the partial band signal added by the calculating means is combined with the signal of the original bandwidth. Band combining means, wherein the partial band signal generating means has zero bit allocation in the partial band signal generating means. As the level of the sub-band signals generated for the partial band, and generates a noise of applying the maximum information separated by said separating means.

A decoder according to claim 18 of the present invention is
The partial band signal storing means for storing the partial band signal of the past frame output from the adding means is provided, and in the partial band signal generating means, the portion of the past frame output from the partial band signal storing means. A band signal is applied.

A decoder according to claim 19 of the present invention is
In the partial band signal generating means, of the partial band signals output from the inverse quantizing means, a composite signal of partial band signals including a harmonic component and a subharmonic component of a target partial band is applied. Is.

[0031]

BEST MODE FOR CARRYING OUT THE INVENTION

First Embodiment of the Invention FIG. 1 shows a coder which is an example of the present invention. In FIG. 1, 1 is a band division unit, 2 is a maximum value detection unit, 3 is an auditory model unit, 4 is a bit allocation unit, 5 is a quantization unit, 6 is a multiplexing unit, and 8 is a reference bit allocation table. It should be noted that 1 to 6 are the same as those in the conventional example.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above-described operation processing of the encoding device, when the partial band signal is quantized and transmitted, the information of the partial band signal with zero bit allocation is not transmitted at all, and therefore the decoder cannot decode and reproduce the partial band signal. . Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. If the bit allocation is 1 or more, information transmission of the partial band signal information is guaranteed,
There is no reduction in power. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. The reference bit allocation table 8 is prepared with reference bit allocation that guarantees power for each partial band in consideration of the power loss when the bit allocation for such partial bands becomes zero. For example, as shown in FIG. 2, at least 1 or more bits are assigned to all partial bands. Using the bit allocation table of FIG. 2 for an input signal having a power distribution as in FIG. 29, transmission of all subband signals is guaranteed,
The transmitted partial band signal is as shown in FIG. 29, and no power loss occurs. Further, when the characteristics such as the power distribution of the input signal are known in advance, such as no high-frequency component at all, or the proportion of the high-frequency component to the signal components in the entire band is extremely small, as shown in FIG. Zero bits are assigned to the high band, and at least one bit is assigned to the other bands. When the bit allocation table of FIG. 3 is used for an input signal having a power distribution as shown in FIG. 29, the subband signal for which transmission is guaranteed is as shown in FIG. Although the power is reduced, the amount of the reduction is small in the total subband signal power and has little influence. In the bit allocation unit 4, first, based on the reference bit allocation table 8, the minimum reference bit allocation for power guarantee is performed in advance. Next, in addition to such minimum standard bit allocation, bit allocation is further performed for each partial band based on the evaluation function output from the auditory model unit 3 in order to improve the quantization distortion. At this time, depending on the value of the evaluation function, one or more bits may be allocated to the partial band whose bit allocation is zero in the reference bit allocation table.
In addition, it will be guaranteed in terms of power. With this, it is possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero. In short, in the encoder configured as described above, the band dividing unit generates the partial band signal by dividing the input signal into a plurality of partial bands, and the maximum value detecting unit determines the absolute value of the partial band signal. Then, the auditory model means performs spectrum analysis on the input signal based on the masking rule of the human auditory characteristics, calculates an evaluation function for a plurality of subbands, and the bit allocation means determines that the bit allocation for the subbands is zero. If the reference bit allocation table with the minimum reference bit allocation prepared in advance is taken into consideration in consideration of the power loss in the event of a loss, the final bit allocation is performed based on the evaluation function, and the quantizing means performs bit allocation. A plurality of sub-band signals are quantized based on, and the multiplexing means multiplexes the bit allocation information, the maximum value detection information and the sample information. To output Te.

Second Embodiment of the Invention FIG. 5 shows an encoder which is an example of the present invention. In FIG. 5, 1 is a band division unit, 2 is a maximum value detection unit, 3 is an auditory model unit, 4 is a bit allocation unit, 5 is a quantization unit, 6 is a multiplexing unit, 9 is a reference bit allocation unit, and 10 is It is a power calculator. It should be noted that 1 to 6 are the same as those in the conventional example.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above operation process of the encoder, when the partial band signal is quantized and transmitted,
Since the information of the partial band signal whose bit allocation is zero is not transmitted at all, the decoder cannot decode and reproduce the partial band signal. Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. When the bit allocation is 1 or more, the information transmission of the partial band signal information is guaranteed and the power is not reduced. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. The reference bit allocation unit 9 is prepared with reference bit allocation for guaranteeing power for each partial band in consideration of power loss when the bit allocation for such partial bands becomes zero. For example, as shown in FIG. 2, at least 1 or more bits are assigned to all partial bands. When the bit allocation table of FIG. 2 is used for an input signal having a power distribution as shown in FIG. 29, transmission of all partial band signals is guaranteed, and the transmitted partial band signals are as shown in FIG. No power loss occurs. Further, when the characteristics such as the power distribution of the input signal are known in advance, such as no high-frequency component at all, or the proportion of the high-frequency component to the signal components in the entire band is extremely small, as shown in FIG. Zero bits are assigned to the high band, and at least one bit is assigned to the other bands. When the bit allocation unit 9 of FIG. 3 is used for the input signal having the power distribution as shown in FIG. 29, the partial band signal for which transmission is guaranteed becomes as shown in FIG.
The power of the sub-band signal with zero bit allocation decreases, but the decrease has a small effect on the total power of the sub-band signals. In the bit allocation unit 4, first, based on the reference bit allocation unit 9, the minimum reference bit allocation for power guarantee is performed in advance. Next, in addition to such minimum standard bit allocation, bit allocation is further performed for each partial band based on the evaluation function output from the auditory model unit 3 in order to improve the quantization distortion. At this time, depending on the value of the evaluation function, one or more bits may be allocated to the partial band whose bit allocation is zero in the reference bit allocation table, and the power is further guaranteed. With this, it is possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero.

Next, the power calculation unit 10 calculates the power from the partial band signal of each partial band output from the band division unit 1 as the power for each partial band. The reference bit allocation unit 9 considers the power loss when the bit allocation to the partial band becomes zero, from the calculation result of the power of each partial band by the power calculation unit 10. The method of consideration is as described above. As a result, for example, at least one bit is assigned to all the partial bands. Further, when it is judged that there is no high frequency component at all or the ratio of the high frequency component to the signal component of the entire band is extremely small, it is zero for the high frequency band and at least 1 for the other frequency bands. Allocate more bits. In the bit allocation unit 4, in addition to the lowest standard bit allocation for power assurance allocated by the standard bit allocation unit 9, each part for improving the quantization distortion based on the evaluation function output from the auditory model unit 3. Further bit allocation is performed for the band. This makes it possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero in response to the change of the input signal. Note that the reference bit allocation unit 9 may select from a plurality of reference bit allocation tables prepared in advance, instead of performing arbitrary reference bit allocation based on the power calculation result. In short, the band dividing means generates the partial band signal by dividing the input signal into a plurality of partial bands, the power calculating means calculates the power of the partial band signal, and the maximum value detecting means calculates the partial band signal. The maximum absolute value is detected, and the auditory model means
The input signal is spectrum-analyzed based on the masking rule of the human auditory characteristics, the evaluation function for a plurality of partial bands is calculated, and the reference bit allocating means outputs the partial information from the power information of the partial band signal output from the band dividing means. Considering the power loss when the bit allocation to the band becomes zero, the lowest standard bit allocation is performed, and the bit allocation means adds the final standard bit allocation based on the evaluation function in addition to the minimum standard bit allocation. Allocation is performed, the quantizing means quantizes the plurality of partial band signals based on the bit allocation, and the multiplexing means multiplexes the bit allocation information, the maximum value detection information and the sample information and outputs the multiplexed information.

Embodiment 3 of the Invention FIG. 6 shows an encoder which is an example of the present invention. In FIG. 6, 1 is a band division unit, 2 is a maximum value detection unit, 3 is an auditory model unit, 4 is a bit allocation unit, 5 is a quantization unit, 6 is a multiplexing unit, and 9 is a reference bit allocation unit. It should be noted that 1 to 6 are the same as those in the conventional example.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above operation process of the encoder, when the partial band signal is quantized and transmitted,
Since the information of the partial band signal whose bit allocation is zero is not transmitted at all, the decoder cannot decode and reproduce the partial band signal. Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. When the bit allocation is 1 or more, the information transmission of the partial band signal information is guaranteed and the power is not reduced. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. The reference bit allocation unit 9 is prepared with reference bit allocation for guaranteeing power for each partial band in consideration of power loss when the bit allocation for such partial bands becomes zero. For example, as shown in FIG. 2, at least 1 or more bits are assigned to all partial bands. When the bit allocation table of FIG. 2 is used for an input signal having a power distribution as shown in FIG. 29, transmission of all partial band signals is guaranteed, and the transmitted partial band signals are as shown in FIG. No power loss occurs. Further, when the characteristics such as the power distribution of the input signal are known in advance, such as no high-frequency component at all, or the proportion of the high-frequency component to the signal components in the entire band is extremely small, as shown in FIG. Zero bits are assigned to the high band, and at least one bit is assigned to the other bands. When the bit allocation unit 9 of FIG. 3 is used for the input signal having the power distribution as shown in FIG. 29, the partial band signal for which transmission is guaranteed becomes as shown in FIG.
The power of the sub-band signal with zero bit allocation is reduced, but the amount of the decrease is small in the total power of the sub-band signal and has little influence. In the bit allocation unit 4, first, based on the reference bit allocation unit 9, the minimum standard bit allocation for power guarantee is performed in advance. next,
In addition to such minimum standard bit allocation, bit allocation is further performed for each partial band based on the evaluation function output from the auditory model unit 3 in order to improve quantization distortion. At this time, depending on the value of the evaluation function, one or more bits may be allocated to the partial band whose bit allocation is zero in the reference bit allocation table, and the power is further guaranteed. With this, it is possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero.

Next, the reference bit allocation unit 9 will be described in detail. The reference bit allocation unit 9 calculates the power of each partial band from the result of the spectrum analysis in the auditory model unit 3, and considers the power loss when the bit allocation to the partial band becomes zero. The method of consideration is as described above. As a result, for example, at least one bit is assigned to all the partial bands. Further, when it is judged that there is no high frequency component at all or the ratio of the high frequency component to the signal component of the entire band is extremely small, it is zero for the high frequency band and at least 1 for the other frequency bands. Allocate more bits. In the bit allocation unit 4, in addition to the minimum standard bit allocation for power guarantee allocated by the standard bit allocation unit 9,
Based on the evaluation function output from the auditory model unit 3, bits are further allocated to each subband in order to improve quantization distortion. This makes it possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero in response to the change of the input signal. It should be noted that the reference bit allocation unit 9 may select from a plurality of reference bit allocation tables prepared in advance as described above, instead of performing arbitrary reference bit allocation. In short, the band dividing means generates the partial band signal by dividing the input signal into a plurality of partial bands, the power calculating means calculates the power of the partial band signal, and the maximum value detecting means detects the partial band signal. The maximum value of the absolute value is detected, the auditory model means performs spectrum analysis of the input signal based on the masking rule of the human auditory characteristics, calculates the evaluation function for a plurality of sub-bands, and the reference bit assigning means determines the auditory model. From the spectrum analysis result by the means, the lowest standard bit allocation is performed in consideration of the power loss when the bit allocation to the partial band becomes zero, and the bit allocation means evaluates in addition to the minimum standard bit allocation. The final bit allocation is performed based on the function, and the quantizing means quantizes the plurality of sub-band signals based on the bit allocation, It means,
The bit allocation information, the maximum value detection information, and the sample information are multiplexed and output.

Embodiment 4 of the Invention FIG. 7 shows a coder which is an embodiment of the present invention. In FIG. 7, 1 to 6 are the same as the above-mentioned conventional example, and the description thereof is omitted. Reference numeral 9 is a reference bit allocation unit.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above operation process of the encoder, when the partial band signal is quantized and transmitted,
Since the information of the partial band signal whose bit allocation is zero is not transmitted at all, the decoder cannot decode and reproduce the partial band signal. Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. When the bit allocation is 1 or more, the information transmission of the partial band signal information is guaranteed and the power is not reduced. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. The reference bit allocation unit 9 is prepared with reference bit allocation for guaranteeing power for each partial band in consideration of power loss when the bit allocation for such partial bands becomes zero. For example, as shown in FIG. 2, at least 1 or more bits are assigned to all partial bands. When the bit allocator of FIG. 2 is used for an input signal having a power distribution as shown in FIG. 29, transmission of all partial band signals is guaranteed, and the transmitted partial band signals are as shown in FIG. No power loss occurs. Further, when the characteristics such as the power distribution of the input signal are known in advance, such as no high-frequency component at all, or the proportion of the high-frequency component to the signal components in the entire band is extremely small, as shown in FIG. Zero for high frequencies,
At least 1 or more bits are assigned to other bands. When the bit allocation table of FIG. 3 is used for an input signal having a power distribution as shown in FIG. 29, the subband signal for which transmission is guaranteed is as shown in FIG. Power is reduced,
The amount of the reduction is small in the total power of the partial band signal and has little influence. In the bit allocation unit 4, first, based on the reference bit allocation unit 9, the minimum standard bit allocation for power guarantee is performed in advance. Next, in addition to such minimum standard bit allocation, bit allocation is further performed for each partial band based on the evaluation function output from the auditory model unit 3 in order to improve the quantization distortion. At this time, depending on the value of the evaluation function, one or more bits may be allocated to the partial band whose bit allocation is zero in the reference bit allocation table, and the power is further guaranteed. With this, it is possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero.

Next, the reference bit allocation section 9 will be described in detail. The reference bit allocation unit 9 calculates the power of each partial band from the maximum absolute value of the sample data of each partial band output from the maximum value detection unit 2, and when the bit allocation to the partial band becomes zero. Consider power loss. The method of consideration is as described above. As a result, for example, at least one bit is assigned to all the partial bands. Further, when it is judged that there is no high frequency component at all or the ratio of the high frequency component to the signal component of the entire band is extremely small, it is zero for the high frequency band and at least 1 for the other frequency bands. Allocate more bits. In the bit allocation unit 4, in addition to the lowest standard bit allocation for power assurance allocated by the standard bit allocation unit 9, each part for improving the quantization distortion based on the evaluation function output from the auditory model unit 3. Further bit allocation is performed for the band. This makes it possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero in response to the change of the input signal. Note that the reference bit allocation unit 9 may select from a plurality of reference bit allocation tables prepared in advance, instead of performing arbitrary reference bit allocation based on the power calculation result. In short, the band dividing means generates the partial band signal by dividing the input signal into a plurality of partial bands, the power calculating means calculates the power of the partial band signal, and the maximum value detecting means calculates the partial band signal. The maximum value of the absolute value is detected, the auditory model means performs spectrum analysis of the input signal based on the masking rule of human auditory characteristics, calculates the evaluation function for a plurality of partial bands, and the reference bit allocation means determines the maximum value. Considering the power loss when the bit allocation to the partial band becomes zero from the maximum value in the detection means, the lowest standard bit allocation is performed, and the bit allocation means adds the minimum standard bit allocation. ,
The final bit allocation is performed based on the evaluation function, the quantizing means quantizes the plurality of partial band signals based on the bit allocation, and the multiplexing means performs the bit allocation information, the maximum value detection information, and the sample information. Multiplex and output.

Fifth Embodiment of the Invention FIG. 8 shows an encoder which is an embodiment of the present invention. In FIG. 8, 1 to 6 are the same as the above-mentioned conventional example, and the description thereof is omitted. Reference numeral 9 is a reference bit allocation unit, 11 is a local decoding unit, and 12 and 13 are power analysis units.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above operation process of the encoder, when the partial band signal is quantized and transmitted,
Since the information of the partial band signal whose bit allocation is zero is not transmitted at all, the decoder cannot decode and reproduce the partial band signal. Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. When the bit allocation is 1 or more, the information transmission of the partial band signal information is guaranteed and the power is not reduced. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. The reference bit allocation unit 9 is prepared with reference bit allocation for guaranteeing power for each partial band in consideration of power loss when the bit allocation for such partial bands becomes zero. For example, as shown in FIG. 2, at least 1 or more bits are assigned to all partial bands. When the bit allocation unit 9 of FIG. 2 is used for the input signal having the power distribution as shown in FIG.
Transmission of all partial band signals is guaranteed, the transmitted partial band signals are as shown in FIG. 29, and no power loss occurs. Further, when the characteristics such as the power distribution of the input signal are known in advance, such as no high-frequency component at all, or the proportion of the high-frequency component to the signal components in the entire band is extremely small, as shown in FIG. Zero bits are assigned to the high band, and at least one bit is assigned to the other bands. When the bit allocation unit 9 of FIG. 3 is used for an input signal having a power distribution as shown in FIG. 29, the subband signal whose transmission is guaranteed is as shown in FIG. Power is reduced,
The reduced amount has a small influence on the power of the entire partial band signal and has little influence. In the bit allocation unit 4, first, based on the reference bit allocation unit 9, the minimum reference bit allocation for power guarantee is performed in advance. Next, in addition to such a minimum standard bit allocation,
Bit allocation is further performed for each subband in order to improve quantization distortion based on the output evaluation function. At this time, depending on the value of the evaluation function, one or more bits may be allocated to the partial band whose bit allocation is zero in the reference bit allocation table, and the power is further guaranteed. With this, it is possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero.

Next, the bit allocation unit 4 once allocates bits to each partial band based on the evaluation function output from the auditory model unit 3. The quantizer 5 performs quantization based on the bit allocation, and the multiplexer 6 multiplexes various information. The local decoding unit 11 has the same configuration as the decoder shown in the conventional example, and performs the reverse procedure of the processing in the encoder, that is, the operation of separating encoded data, dequantizing, band synthesizing, and the like. The local decoding unit 11 inputs the encoded data output from the multiplexing unit 6, performs the above processing, and outputs a decoded signal. The first power analysis unit 12 calculates the power and spectrum of the input original signal input to the encoder, and the second power analysis unit 13 calculates the power and spectrum of the decoded signal output from the local decoding unit 11. calculate. The reference bit allocation unit 9 includes the first power analysis unit 12
Also, the power loss in encoding is considered from the power analysis result output from the second power analysis unit 13. The method of consideration is as described above. As a result, in order to prevent power loss, for example, at least one bit is assigned to all subbands. Further, when it is judged that there is no high frequency component at all or the ratio of the high frequency component to the signal component of the entire band is extremely small, it is zero for the high frequency band and at least 1 for the other frequency bands. Allocate more bits. In the bit allocation unit 4,
In addition to the lowest standard bit allocation for power assurance allocated by the standard bit allocation unit 9, based on the evaluation function output from the auditory model unit 3, a bit is again added to each subband for improving the quantization distortion. Make an assignment. This makes it possible to prevent the loss of the partial band signal component due to the bit allocation to the partial band becoming zero in response to the change of the input signal. Note that the reference bit allocation unit 9 may select from a plurality of reference bit allocation tables prepared in advance, instead of performing arbitrary reference bit allocation based on the power calculation result. In short, the band dividing means generates a partial band signal by dividing the input signal into a plurality of partial bands, the power calculating means calculates the power of the partial band signal, and the maximum value detecting means determines the partial band signal. The maximum value of the absolute value of the signal is detected, the auditory model means performs the spectrum analysis of the input signal based on the masking rule of the human auditory characteristics, calculates the evaluation function for a plurality of partial bands, and the reference bit allocation means, From the power analysis results of the input signal and the decoded signal, the lowest standard bit allocation is performed in consideration of the power loss when the bit allocation to the partial band becomes zero, and the bit allocation means selects the minimum standard bit allocation. In addition, final bit allocation is performed based on the evaluation function, and the quantizing means quantizes a plurality of partial band signals based on the bit allocation and multiplexes. The means multiplexes and outputs the bit allocation information, the maximum value detection information and the sample information, the local decoder generates a decoded signal, and the first power analysis means calculates the power and spectrum of the input signal. , A second power analysis means calculates the power and spectrum of the decoded signal.

Sixth Embodiment of the Invention FIG. 9 shows an encoder which is an example of the present invention. In FIG. 9, 1 to 6 are the same as the above-mentioned conventional example, and the description thereof is omitted. Reference numeral 14 is a power loss calculation unit, and 15 is a gain adjustment unit.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above operation process of the encoder, when the partial band signal is quantized and transmitted,
Since the information of the partial band signal whose bit allocation is zero is not transmitted at all, the decoder cannot decode and reproduce the partial band signal. Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. When the bit allocation is 1 or more, the information transmission of the partial band signal information is guaranteed and the power is not reduced. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. Next, the band division unit 1, the maximum value detection unit 2,
The auditory model unit 3 and the bit allocation unit 4 once process the input signal. The power loss calculation unit 14 calculates the power of the partial band signal from the partial band signals of the partial bands output from the band division unit 1, and at the same time, performs the bit allocation for each partial band from the bit allocation determined by the bit allocation unit 4. Is determined to be zero, and the total power of the subbands with zero bit allocation is calculated. FIG. 10 shows the power of the partial band signal. The shaded area in the figure represents the power of the partial band in which the bit allocation is zero, and this is the power loss. The gain adjusting unit 15 adjusts the gain of the input signal in a direction of compensating for the loss according to the power calculated by the power loss calculating unit 14. After that, normal processing is performed on the input signal adjusted by the gain adjusting unit 15. Since the input signal in this case is gain-adjusted, the power loss calculation unit 1
The power of the partial band signal at 4 has increased to the level shown by the dotted line in FIG. This dotted line portion is equal to the power of the shaded portion, so that the loss of the partial band signal component due to the zero bit allocation to the partial band can be compensated in advance on the encoder side.
In short, the band dividing means generates the partial band signal by dividing the input signal into a plurality of partial bands, and the maximum value detecting means detects the maximum absolute value of the partial band signal,
The auditory model means performs spectrum analysis of the input signal based on the masking rule of the human auditory characteristics, calculates an evaluation function for a plurality of subbands, and the bit allocation means performs a bit for quantizing the plurality of subband signals. As the allocation, bit allocation is performed based on the evaluation function, the quantizing means quantizes a plurality of partial band signals based on the bit allocation, and the multiplexing means performs bit allocation information, maximum value detection information, and sample information. The multiplexed power is output, and the power loss calculation means calculates the power loss when the bit allocation to the partial band becomes zero from the power of the partial band signal output from the band division means and the bit allocation by the bit allocation means. The gain adjusting unit adjusts the gain of the input signal according to the power loss.

Seventh Embodiment of the Invention FIG. 11 shows an encoder which is an embodiment of the present invention. FIG.
In the above, 1 to 6 are the same as those in the above-mentioned conventional example, and the description thereof will be omitted. Reference numeral 14 is a power loss calculation unit, and 16 is a maximum value adjustment unit. Next, the operation will be described.

Next, the operation will be described. Band division unit 1
Then, the input signal is divided into a plurality of partial bands, and the partial band signal for a certain specific time section is output. In the case of audio signal coding, it is usually divided into 32 equal bandwidths according to the human hearing characteristics. The maximum value detection unit 2 detects the maximum absolute value of the partial band signal for each partial band within the specific time segment. The auditory model unit 3 performs spectrum analysis such as FFT analysis on the input signal, further analyzes it based on human auditory characteristics, and considers the maximum value from the maximum value detection unit 2 to consider the bit allocation unit. In 4, the evaluation function for optimal bit allocation is calculated. The human auditory characteristics mentioned here are mainly the minimum audible limit and the masking effect. The minimum audible limit is the minimum level that can be perceived by human hearing, and the masking effect is a phenomenon in which a small level signal cannot be perceived by a large level signal. Considering these characteristics, the evaluation function is calculated from the relationship between the spectrum of the input signal component and its mask characteristics. An example of the evaluation function is the difference between the maximum value of the signal level and the minimum value of the mask characteristic in each partial band. The bit allocation unit 4 determines the optimum bit allocation for each partial band based on the evaluation function from the auditory model unit 3. The quantizing unit 5 normalizes the partial band signals of the respective partial bands from the band dividing unit 1 with the maximum value from the maximum value detecting unit 2 in order to improve the quantization efficiency, and quantizes them according to the bit allocation from the bit allocating unit 4. And encode. The multiplexing unit 6 multiplexes the maximum value information from the maximum value detection unit 2, the bit allocation information from the bit allocation unit 4, and the sample information from the quantization unit 5, and outputs it as encoded data. At this time, in order to reduce the amount of information, when the bit allocation to the partial band is zero, the maximum value information and the sample information for the partial band are usually not multiplexed. In the above operation process of the encoder, when the partial band signal is quantized and transmitted,
Since the information of the partial band signal whose bit allocation is zero is not transmitted at all, the decoder cannot decode and reproduce the partial band signal. Therefore, in the reproduced decoded signal, the power of the partial band signal is reduced with respect to the power of the input signal on the encoder side. When the bit allocation is 1 or more, the information transmission of the partial band signal information is guaranteed and the power is not reduced. Furthermore, if the bit allocation is at least 1 bit, it is possible to prevent the power reduction, and the bit allocation more than that contributes to the improvement of the quantization distortion. The power loss calculation unit 14 includes the band division unit 1
The power of the sub-band signal is calculated from the sub-band signal of each sub-band output from, and at the same time, it is determined from the bit allocation determined by the bit allocation unit 4 whether the bit allocation to each sub-band is zero, Calculate the sum of the powers of the subbands with zero bit allocation. This is the power loss due to zero bit allocation. The maximum value adjusting unit 16 adjusts the maximum value of gain for all the partial bands detected by the maximum value detecting unit 2 in the direction of compensating for the power loss calculated by the power loss calculating unit 14. . This gain adjustment may be the same for all the partial bands or may be the gain weighted. The quantization unit 5 performs normalization using the maximum value output from the maximum value detection unit 2, and the maximum value information output from the maximum value adjustment unit 16 is used as the maximum value information to be transmitted. Therefore, the gain is given to the partial band signal when the inverse normalization is performed on the decoder side. This is shown in FIG. After normalizing a certain sub-band signal with the maximum value, the gain is adjusted to the maximum value and inverse normalization is performed to amplify the amplitude, and as a result, the power of the sub-band signal also increases. This action compensates for the power loss due to zero bit allocation by increasing the power of all subbands. As a result, the loss of the partial band signal component due to the zero bit allocation to the partial band can be compensated in advance on the encoder side. In short, the band dividing means generates the partial band signal by dividing the input signal into a plurality of partial bands, and the maximum value detecting means detects the maximum absolute value of the partial band signal, and the auditory model means. Is the spectrum analysis of the input signal based on the masking rule of human auditory characteristics, calculates the evaluation function for a plurality of subbands, the bit allocation means, as a bit allocation for quantizing the plurality of subband signals, Bit allocation is performed based on the evaluation function, the quantizing means quantizes the plurality of partial band signals based on the bit allocation, and the multiplexing means multiplexes the bit allocation information, the maximum value detection information, and the sample information. The power loss calculation means outputs the partial band signal from the power of the partial band signal output from the band division means and the bit allocation by the bit allocation means. Against calculates the power loss when the bit allocation becomes zero, adjust the maximum value.

Eighth Embodiment of the Invention In the seventh embodiment, the maximum value adjusting unit 16 compensates the power loss calculated by the power loss calculating unit 14 according to the power loss, and the maximum value adjusting unit 16 detects the maximum value for each partial band detected by the maximum value detecting unit 2. Among the values, the gain adjustment is performed only for the largest value. By doing so, as described in the seventh embodiment, the power loss due to the zero bit allocation is compensated by increasing the power of the partial band having the largest maximum value. As a result, the loss of the partial band signal component due to the zero bit allocation to the partial band can be compensated in advance on the encoder side. Further, in the maximum value adjusting unit 16, as a method of selecting a target subband for which the maximum value gain adjustment is performed, among the partial bands calculated by the power loss calculating unit 14, the partial band having the largest power of the partial band signal is selected. It may be selected. Further, in the maximum value adjusting unit 16, as a method of selecting the target subband for which the gain adjustment of the maximum value is performed, the partial band may be determined by selecting from the evaluation function output from the auditory model unit 3. When the difference between the maximum value of the signal level and the minimum value of the mask characteristic is considered as the evaluation function in the conventional example, the smallest partial band is selected. Further, in the above-described embodiment, the gain adjustment of the maximum value is not performed for only one partial band, but may be performed for the N highest partial bands, for example.

Ninth Embodiment of the Invention FIG. 13 shows an encoder which is an example of the present invention. In FIG. 13, 1 to 6 are the same as the above-mentioned conventional example, and the description thereof is omitted. Reference numeral 14 is a power loss calculation unit, and 15 is a gain adjustment unit. Next, the operation will be described.

In the above sixth to eighth embodiments, the power loss calculating section 14 calculates the power of each partial band signal from the spectrum analysis result such as FFT performed in the auditory model section 3 as a method of calculating the power for each partial band. It may be one that does. The description will be specifically given based on the sixth embodiment. FIG. 13 shows an encoder which is an example of the present invention.

Next, the operation processing will be described. The band division unit 1, the maximum value detection unit 2, the auditory model unit 3, and the bit allocation unit 4 once process the input signal. The power loss calculation unit 14 calculates the total power of the partial band in which the bit allocation is zero, from the spectrum analysis result in the auditory model unit 3 and the bit allocation determined by the bit allocation unit 4. The gain adjusting unit 15 adjusts the gain of the input signal according to the power loss calculated by the power loss calculating unit 14. Normal processing is performed on the input signal adjusted by the gain adjusting unit 15. by this,
The loss of the partial band signal component due to the zero bit allocation to the partial band can be compensated in advance. Further, the gain adjusting unit 15 may perform the gain adjustment as in the seventh and eighth embodiments.

Tenth Embodiment of the Invention FIG. 14 shows an encoder which is an example of the present invention. In FIG. 14, 1 to 6 are the same as the above-mentioned conventional example, and the description thereof is omitted. Reference numeral 14 is a power loss calculation unit, and 15 is a gain adjustment unit. Next, the operation will be described.

The band division unit 1, the maximum value detection unit 2, the auditory model unit 3, and the bit allocation unit 4 once process the input signal. The power loss calculation unit 14 calculates the total power of the partial band in which the bit allocation is zero, from the maximum value detected by the maximum value detection unit 2 and the bit allocation determined by the bit allocation unit 4. Gain adjuster 1
Reference numeral 5 adjusts the gain of the input signal according to the power loss calculated by the power loss calculating unit 14. Normal processing is performed on the input signal adjusted by the gain adjusting unit 15. As a result, the loss of the partial band signal component due to the zero bit allocation to the partial band can be compensated in advance. In addition, the gain adjusting unit 15
In, the gain adjustment as in the seventh and eighth embodiments may be performed.

Eleventh Embodiment of the Invention FIG. 15 shows an encoder which is an example of the present invention. In FIG. 15, 1 to 6 are the same as those in the above-mentioned conventional example, and the description thereof is omitted. Reference numeral 11 is a local decoding unit, 12 and 13 are power analyzing units, 14 is a power loss calculating unit, and 15 is a gain adjusting unit. Next, the operation will be described.

In the above sixth to eighth embodiments, the power loss calculating section 14 calculates the power of each partial band signal from the maximum value output from the maximum value detecting section 2 as a method of calculating the power for each partial band. It may be one. The description will be specifically given based on the sixth embodiment.
FIG. 15 shows an encoder which is an example of the present invention.

The local decoding unit 11 has the same configuration as the decoder shown in the conventional example and performs the same operation. Local decoding unit 11
Receives the decoded data output from the multiplexing unit 6 and outputs a decoded signal. The first power analysis unit 12 calculates the power of the input original signal input to the encoder, and the second power analysis unit 13 calculates the power of the decoded signal output from the local decoding unit 11. The power loss calculation unit 14 calculates the power loss from the power analysis results output from the first power analysis unit 12 and the second power analysis unit 13. The gain adjusting unit 15 adjusts the gain according to the power loss output from the power loss calculating unit 14. by this,
The power loss when the bit allocation becomes zero can be compensated in advance by manipulating the power of the input signal. Further, the gain adjusting unit 15 may perform the gain adjustment as in the seventh and eighth embodiments.

Example 1. FIG. 16 shows an encoder which is an example of the present invention. In FIG. 16, 1 to
6 is the same as the above-mentioned conventional example, and its description is omitted. 12 is a power analysis unit. Next, the operation will be described.

The power analysis unit 12 calculates the power of the input original signal and outputs the information as power information. The multiplexing unit 6 multiplexes and encodes the bit allocation information from the bit allocation unit 4, the maximum value information from the maximum value detection unit 2, the sample information from the quantization unit 5, and the power information from the power analysis unit 12 into a code. Output as the converted data.

Twelfth Embodiment of the Invention FIG. 17 shows a decoder which is an example of the present invention. In FIG. 17, reference numerals 21 to 23 are the same as those in the conventional example,
The description is omitted. Reference numeral 15 is a gain adjusting unit. Next, the operation will be described.

The separating section 21 separates the encoded data into bit allocation information, maximum value information, sample information and power information. The power information mentioned here is the power of the input original signal in the encoder. The inverse quantizing unit 22 decodes the partial band signal from the above information, but for the partial band whose bit allocation is zero, the partial band signal cannot be decoded and the partial band signal becomes zero. Therefore, when the subband signal originally existed in the encoder, the decoded signal output from the band synthesis unit 23 is reduced by the power of the subband signal whose bit allocation is zero.
In order to compensate for this decrease in power, the gain adjusting unit 15
While calculating the power of the decoded signal output from the band synthesizer 23, the gain is adjusted so that this power matches the power information of the input original signal in the encoder output from the separator. As a result, the power of the decoded signal can be made equal to that of the original signal. In short, in the decoder configured as described above, the separating means separates the power information, the bit allocation information, the maximum value information and the sample information from the input coded data, and the dequantizing means separates the bit allocation information, the maximum information. The partial band signal is decoded from the value information and the sample information, and the band synthesizing unit band-synthesizes the partial band signal,
The gain adjusting means adjusts the gain of the decoded signal band-combined according to the power information and outputs the decoded signal.

Embodiment 13 of the Invention FIG. 18 shows a decoder which is an example of the present invention. In FIG. 18, 21 to 23 are the same as those in the conventional example,
The description is omitted. Reference numeral 14 is a power loss calculation unit, and 16 is a maximum value adjustment unit. Next, the operation will be described.

The separating unit 21 separates the encoded data into bit allocation information, maximum value information, sample information, and power information. The power information mentioned here is the power of the input original signal in the encoder. The power loss calculation unit 14
First, a partial band whose bit allocation is not zero is discriminated from this bit allocation information, and the power of the partial band signal for that partial band is calculated from the maximum value information for that partial band. Next, the power of the partial band in which the bit allocation is zero is calculated from this and the power information from the separating unit 21. This power becomes a loss of power with respect to the original signal. The maximum value adjusting unit 16 adjusts the maximum value of gain for all partial bands from the separating unit 21 in the direction of compensating for the power loss calculated by the power loss calculating unit 14.
Since the inverse quantization unit 22 performs inverse normalization using the gain-adjusted maximum value, the power of the partial band signal is adjusted, and as a result, the decoded signal output from the band synthesis unit 23 is also adjusted in power. To be done. Thereby, the power of the decoded signal can be compensated. Note that the maximum value adjusting unit 16 may perform gain adjustment only on the largest value of the maximum value information for the partial band output from the separating unit 21.

Example 2. FIG. 19 shows a decoder which is an example of the present invention. In FIG. 19, 21
23 to 23 are the same as those in the above-mentioned conventional example, the description thereof will be omitted. Reference numeral 14 is a power loss calculation unit, and 15 is a gain adjustment unit. Next, the operation will be described.

The separating unit 21 separates the encoded data into bit allocation information, maximum value information, sample information, and power information. The power information mentioned here is the power of the input original signal in the encoder. The power loss calculation unit 14
The power of the partial band signal output from the inverse quantization unit 22 is calculated. In the dequantization unit 22, the subband signal of the subband whose bit allocation is zero is not decoded, and when the subband signal exists in the incoming encoder,
The decoded signal output from the band synthesizing unit 23 is reduced by the power of the partial band signal whose bit allocation is zero. In order to compensate for this decrease in power, the power loss is calculated from the power of the partial band signal output from the dequantization unit 22 calculated previously and the power information from the separation unit 21. The gain adjustment unit 15 adjusts the gain of the partial band signal output from the inverse quantization unit 22 according to the power loss calculated by the power loss calculation unit 14. As a result, the power of the decoded signal output from the band synthesis unit 23 is also adjusted. Thereby, the power of the decoded signal can be compensated. Note that the gain adjustment may be performed on the decoded signal after band synthesis.

Fourteenth Embodiment of the Invention FIG. 20 shows a decoder which is an example of the present invention. In FIG. 20, reference numerals 21 to 23 are the same as those in the conventional example,
The description is omitted. 24 is a partial band signal generator, 25
Is an adder. Next, the operation will be described.

The sub-band signal generator 24 first detects the separation unit 2
From the bit allocation information output from 1, the partial band in which the bit allocation is zero is determined. Next, an alternative signal for the subband is generated. For example, white noise is generated as the alternative signal. At this time, the level of the substitute signal is made equal to the level of the minimum audible limit. This level means the power level of the partial band signal. This minimum audible limit is a limit that can be perceived by human hearing, and when the level is set in this way, this alternative signal component is not perceived even after band synthesis. The addition unit 25 adds the partial band signal decoded by the inverse quantization unit 22 and the partial band signal generated by the partial band signal generation unit as shown in FIG.
By this addition, the partial band signals for all the partial bands are input to the band synthesizing unit 23. In the band synthesizer 23,
Combining the partial band signals output from the adder 25,
Output the decoded signal in the original bandwidth. This makes it possible to compensate the power of the decoded signal.

Fifteenth Embodiment of the Invention FIG. 22 shows a decoder which is an example of the present invention. In FIG. 22, reference numerals 21 to 23 are the same as those in the conventional example,
The description is omitted. 24 is a partial band signal generator, 25
Is an adder. Next, the operation will be described.

The sub-band signal generator 24 first detects the separator 2
From the bit allocation information output from 1, the partial band in which the bit allocation is zero is determined. Next, an alternative signal for the subband is generated. For example, white noise is generated as the alternative signal. At this time, the level of the alternative signal is specified as follows. Based on the bit allocation information output from the separating unit 21, the partial band in which the bit allocation is zero and the partial band in which the bit allocation is not zero are determined. Next, the maximum level information is used to calculate the signal level of the partial band whose bit allocation is not zero. The maximum band information does not exist for the signal level of the partial band in which the bit allocation is zero, and it is obtained as follows. As shown in FIG. 23, the partial band signal of the partial band whose bit allocation is not zero is regarded as a sound that makes other sounds inaudible due to the masking effect, that is, a masker, and the previously obtained partial band whose bit allocation is not zero is considered. From the signal level of, the masking threshold for the subband with zero bit allocation is determined. This masking threshold is used as the level of the alternative signal. If you set the level like this,
This alternative signal component is not perceived even after band synthesis. In addition section 25, as described in the fourteenth embodiment,
The partial band signal decoded by the inverse quantization unit 22 and the partial band signal generated by the partial band signal generation unit are added. By this addition, the partial band signals for all the partial bands are input to the band synthesizing unit 23. The band synthesizing unit 23 synthesizes the partial band signals output from the adding unit 25 and outputs the decoded signal having the original bandwidth. This makes it possible to compensate the power of the decoded signal. As a method of obtaining the signal level of the partial band in which the bit allocation is not zero, it may be obtained from the decoded partial band signal output from the inverse quantization unit 22.

Sixteenth Embodiment of the Invention FIG. 24 shows a decoder which is an example of the present invention. In FIG. 24, 21 to 23 are the same as those in the conventional example,
The description is omitted. 24 is a partial band signal generator, 25
Is an addition unit, and 26 is a level information storage unit. Next, the operation will be described.

The level information storage unit 26 stores the level information of each partial band signal in the frame output from the adding unit 25, and also outputs the level information of each partial band signal in the past frame. The partial band signal generation unit 24 first determines the partial band in which the bit allocation is zero from the bit allocation information output from the separation unit 21.
Next, an alternative signal for the subband is generated. For example, white noise is generated as the alternative signal. At this time, the level of the alternative signal is specified as follows. For example, the level information of the immediately preceding frame for each partial band stored in the level information storage unit 26 is applied as it is. Alternatively, the level value predicted from the transition of the level information of a plurality of past frames of each partial band stored in the level information storage unit 26 is applied. For example, if the level information for the immediately preceding two frames is increasing, the level information is set to 1.2 times the previous level information, and if it is decreasing, it is set to 0.8 times the previous level information. As described in the above embodiment, the adder 25 adds the partial band signal decoded by the inverse quantizer 22 and the partial band signal generated by the partial band signal generator.
By this addition, the partial band signals for all the partial bands are input to the band synthesizing unit 23. In the band synthesizer 23,
Combining the partial band signals output from the adder 25,
Output the decoded signal in the original bandwidth. This makes it possible to compensate the power of the decoded signal.

Seventeenth Embodiment of the Invention FIG. 22 shows a decoder which is an example of the present invention. In FIG. 22, reference numerals 21 to 23 are the same as those in the conventional example,
The description is omitted. 24 is a partial band signal generator, 25
Is an adder. Next, the operation will be described.

The maximum value information for each partial band separated by the separating unit 21 always includes the maximum value for all partial bands regardless of whether the bit allocation to each partial band is zero or non-zero. To do. The partial band signal generation unit 24 first determines the partial band in which the bit allocation is zero from the bit allocation information output from the separation unit 21. Next, an alternative signal for the subband is generated. For example, white noise is generated as the alternative signal. At this time, the level of the alternative signal is specified as follows. The signal level is calculated from the maximum value for this partial band and that level is applied. As described in the above embodiment, the adder 25 adds the partial band signal decoded by the inverse quantizer 22 and the partial band signal generated by the partial band signal generator. By this addition, the partial band signals for all the partial bands are input to the band synthesizing unit 23. Band synthesizer 23
Then, the partial band signals output from the adder 25 are combined to output a decoded signal having the original bandwidth. This makes it possible to compensate the power of the decoded signal.

Eighteenth Embodiment of the Invention FIG. 25 shows a decoder which is an example of the present invention. In FIG. 25, 21 to 23 are the same as those in the conventional example,
The description is omitted. 24 is a partial band signal generator, 25
Is an adding unit, and 26 is a partial band signal storage unit. Next, the operation will be described.

The partial band signal storage unit 26 stores each partial band signal in the frame output from the adding unit 25, and also outputs each partial band signal in the past frame. The partial band signal generation unit 24 first determines the partial band in which the bit allocation is zero from the bit allocation information output from the separation unit 21. Next, an alternative signal for the subband is generated. As the alternative signal, the partial band signal of the previous frame output from the partial band signal storage unit 26 is used. At this time, the level is determined as in the fourteenth to seventeenth embodiments. As described in the seventeenth embodiment, the addition unit 25 adds the partial band signal decoded by the inverse quantization unit 2 and the partial band signal generated by the partial band signal generation unit. By this addition, the partial band signals for all the partial bands are input to the band synthesizing unit 23. Band synthesizer 2
In 3, the partial band signals output from the adder 25 are combined and the decoded signal having the original bandwidth is output. This makes it possible to compensate the power of the decoded signal while maintaining the quality of the decoded signal.

Nineteenth Embodiment of the Invention FIG. 26 shows a decoder which is an example of the present invention. In FIG. 26, 21 to 23 are the same as those in the conventional example,
The description is omitted. 24 is a partial band signal generator, 25
Is an adder. Next, the operation will be described.

The sub-band signal generator 24 first detects the separation unit 2
From the bit allocation information output from 1, the partial band in which the bit allocation is zero is determined. Next, an alternative signal for the subband is generated. A signal generated as follows is used as this alternative signal. In a subband where the bit allocation is zero, the frequency representative of the subband, for example, the center frequency is regarded as the fundamental frequency, and a composite signal is generated from the subband signal of the subband including harmonics and subharmonics to this fundamental frequency. . Here, as a method of selecting the partial band including the harmonic and the subharmonic, for example, a partial band including the harmonic and the subharmonic up to the nth order is selected, or when the fundamental frequency is low, the harmonics up to the nth order are selected. There are methods such as selecting a wave, or selecting subbands including subharmonics up to the nth order when the fundamental frequency is high. The level of the synthesized signal generated in this way is determined as in the above-mentioned fourteenth to seventeenth embodiments. The adder 25 adds the partial band signal decoded by the inverse quantizer 22 and the partial band signal generated by the partial band signal generator. By this addition, the partial band signals for all the partial bands are input to the band synthesizing unit 23. In the band synthesis unit 23, the addition unit 25
The partial band signals output from are combined and the decoded signal of the original bandwidth is output. This makes it possible to compensate the power of the decoded signal while maintaining the quality of the decoded signal. Further, as a method of generating a combined signal from the other partial band signals as described above, a method of combining and generating partial band signals adjacent to the partial band to be generated may be used.

[0078]

As described above, according to the present invention, it is possible to provide a coding / decoding device that obtains a code and a decoded signal without signal power loss from an original signal.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an encoder according to a first embodiment.

FIG. 2 is a diagram showing a storage state of a bit allocation table in the first to fourth embodiments.

FIG. 3 is a diagram showing a storage state of a bit allocation table in the first to fourth embodiments.

FIG. 4 is a diagram showing states of partial band signals in the first to fourth embodiments.

FIG. 5 is a block diagram showing a configuration of an encoder in the second embodiment.

FIG. 6 is a block diagram showing a configuration of an encoder in the third embodiment.

FIG. 7 is a block diagram showing a configuration of an encoder in the fourth embodiment.

FIG. 8 is a block diagram showing a configuration of an encoder in the fifth embodiment.

FIG. 9 is a block diagram showing the configuration of an encoder in the sixth embodiment.

FIG. 10 is a state diagram of a partial band signal according to the sixth embodiment.

FIG. 11 is a block diagram showing a configuration of an encoder according to a seventh embodiment.

FIG. 12 is a state diagram of a partial band signal according to the seventh embodiment.

FIG. 13 is a block diagram showing the configuration of an encoder in the ninth embodiment.

FIG. 14 is a block diagram showing a configuration of an encoder in the tenth embodiment.

FIG. 15 is a block diagram showing a configuration of an encoder according to the eleventh embodiment.

FIG. 16 is a block diagram showing a configuration of an encoder in the first embodiment.

FIG. 17 is a block diagram showing the configuration of the decoder in the twelfth embodiment.

FIG. 18 is a block diagram showing the structure of the decoder in the thirteenth embodiment.

FIG. 19 is a block diagram showing a configuration of a decoder in the second embodiment.

FIG. 20 is a block diagram showing the configuration of the decoder according to the fourteenth embodiment.

FIG. 21 is a state diagram of a partial band signal in the fourteenth embodiment.

FIG. 22 is a block diagram showing the configuration of the decoder in the fifteenth embodiment.

FIG. 23 is a state diagram of a partial band signal according to the fifteenth embodiment.

FIG. 24 is a block diagram showing the configuration of the decoder in the sixteenth embodiment.

FIG. 25 is a block diagram showing the structure of the decoder according to the seventeenth embodiment.

FIG. 26 is a block diagram showing the structure of the decoder in the eighteenth embodiment.

FIG. 27 is a block diagram showing a configuration of an encoder in a conventional example.

FIG. 28 is a diagram showing a power analysis state of an input signal in the conventional example.

FIG. 29 is a diagram showing a state of bit allocation in the conventional example.

FIG. 30 is a diagram showing a power analysis state of a partial band signal in a conventional example.

FIG. 31 is a block diagram showing a configuration of a decoder in a conventional example.

[Explanation of symbols]

1 band dividing means, 2 maximum value detecting means, 3 auditory model means, 4 bit allocating means, 5 quantizing means, 6
Multiplexing means, 8 reference bit allocation table, 9 reference bit allocation means, 10 power calculation means, 11 local decoding means, 12 first power analysis means, 13 second power analysis means, 14 power loss calculation means, 15 gain adjustment means , 16 maximum value adjusting means, 21 separating means, 22
Inverse quantization means, 23 band synthesis means, 24 partial band signal generation means, 25 addition section, 26 level information storage section, 27 partial band signal storage section.

Front page continuation (72) Inventor Noriaki Kono 2-3-3 Marunouchi, Chiyoda-ku, Tokyo Sanryo Electric Co., Ltd. (72) Inventor Yuji Naito 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Sanryo Denki Within the corporation

Claims (19)

[Claims] 1. A band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and spectrum analysis of the input signal based on a masking rule of human auditory characteristics, and Auditory model means for calculating an evaluation function for a partial band, a reference bit allocation table for storing predetermined reference bit allocation information for the partial band, reference bit allocation information of the reference bit allocation table, and the auditory model means. Based on the evaluation function calculated in
Bit allocation means for generating bit allocation information for quantizing the plurality of partial band signals generated by the band division means, and generation by the band division means based on the bit allocation information generated by the bit allocation means And a quantizing means for quantizing the plurality of generated partial band signals. 2. A band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a power calculating means for calculating the power of the plurality of partial band signals generated by the band dividing means. A spectrum analysis of the input signal based on a masking rule of human auditory characteristics, and an auditory model means for calculating an evaluation function for the plurality of partial bands; A reference bit allocation table for outputting predetermined reference bit allocation information for a partial band, and bit allocation information for quantizing a plurality of partial band signals generated by the band dividing means are defined in the reference bit allocation table. Bit allocation means for generating based on the bit allocation information and the evaluation function calculated by the auditory model means, Based on the bit allocation information determined by the bit allocation means, the encoder characterized in that a plurality of partial band signals generated by said band dividing means and a quantization means for quantizing. 3. Band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and spectrum analysis of the input signal based on a masking rule of human auditory characteristics, and Based on the auditory model means for calculating the evaluation function for the sub-band, and the spectrum analysis result of the auditory model means,
A reference bit allocation table for outputting predetermined reference bit allocation information for the sub-bands, and a bit allocation for quantizing the plurality of sub-band signals generated by the band dividing means, the reference bit allocation information of the reference bit allocation table. And a bit allocation unit that is generated based on the evaluation function calculated by the auditory model unit, and a plurality of partial band signals that are generated by the band division unit based on the bit allocation information determined by the bit allocation unit. And a quantization means for performing the encoding. 4. A band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a maximum value for detecting a maximum value of the plurality of partial band signals generated by the band dividing means. Detection means, spectrum analysis of the input signal based on a masking rule of human auditory characteristics, auditory model means for calculating an evaluation function for the plurality of partial bands, and maximum value detected by the maximum value detection means based on,
A reference bit allocation table for storing predetermined reference bit allocation information for a partial band, and a bit allocation for quantizing a plurality of partial band signals generated by the band dividing means, the reference bit allocation information of the reference bit allocation table. And a bit allocation unit which is generated based on the evaluation function calculated by the auditory model unit, and a plurality of partial band signals which are generated by the band division unit are quantized based on the bit allocation information determined by the bit allocation unit. And a quantization means for performing the encoding. 5. A band dividing means for dividing an input signal into a plurality of partial band signals to generate a plurality of partial band signals, and a spectrum analysis of the input signal based on a masking rule of human auditory characteristics, and a plurality of the plurality of partial band signals. A hearing model means for calculating an evaluation function for the partial band, a reference bit allocation table for storing predetermined reference bit allocation information for the partial band, and a plurality of partial band signals generated by the band dividing means for quantizing Based on the reference bit allocation information of the reference bit allocation table and the evaluation function calculated by the auditory model means, and the band allocation information based on the bit allocation information determined by the bit allocation means. Quantizing means for quantizing a plurality of partial band signals generated by the dividing means, and A local decoding means for decoding the quantized encoded data by the quantization means, first to calculate the power and spectrum of the input signal
Power analyzing means and second power analyzing means for calculating the power and spectrum of the decoded signal decoded by the local decoding means, and the reference bit allocating means includes the first power analyzing means and the first power analyzing means. An encoder which outputs predetermined reference bit allocation information in consideration of power loss when bit allocation to a partial band is reduced, as a result of the power analysis means of No. 2. 6. The input signal is divided into a plurality of partial bands,
Band division means for generating a plurality of partial band signals; auditory model means for spectrum analysis of the input signal based on a masking rule of human auditory characteristics to calculate an evaluation function for the plurality of partial bands; Bit allocation means for generating bit allocation information for quantizing the plurality of partial band signals generated by the plurality of sub-band signals, and a plurality of bit allocation information generated by the band division means based on the bit allocation information determined by the bit allocation means. Quantizing means for quantizing the partial band signal, power calculating means for obtaining the power of the partial band signal of the band dividing means, and gain of the input signal corresponding to the partial band power obtained by the power calculating means An encoder provided with adjusting means for adjusting. 7. The input signal is divided into a plurality of partial bands,
Band dividing means for generating a plurality of partial band signals, maximum value detecting means for detecting a maximum absolute value of the plurality of partial band signals generated by the band dividing means, and masking an input signal for human auditory characteristics Auditory model means for performing spectrum analysis based on rules to calculate an evaluation function for the plurality of partial bands, and bits for generating bit allocation information for quantizing the plurality of partial band signals generated by the band dividing means. Allocating means, quantizing means for quantizing the plurality of partial band signals generated by the band dividing means based on the bit allocating information determined by the bit allocating means, and power of the partial band signal of the band dividing means And the maximum value detection corresponding to the power of the partial band obtained by the power calculation means. Encoder is characterized in that an adjusting means for adjusting the gain of the detected maximum value means. 8. The maximum value adjusting means, based on the bit allocation information determined by the bit allocation means and the power information of the partial band calculated by the power calculating means, has zero bit allocation to the partial band. 8. The encoder according to claim 7, wherein the gain adjustment of the largest maximum value among the maximum values for each partial band detected by said maximum value detecting means is performed in consideration of the power loss in the case. 9. The power calculation means considers the power loss when the bit allocation to the partial band becomes zero from the bit allocation information determined by the bit allocation means and the spectrum analysis result in the auditory model means. The encoder according to claim 6 or 8, characterized by: 10. The power calculation means, when the bit allocation to the partial band becomes zero from the bit allocation determined by the bit allocation means and the maximum value for the partial band detected by the maximum value detection means. 9. The encoder according to claim 6, wherein the power loss of the encoder is taken into consideration. 11. A local decoding means for decoding the coded data generated by the multiplexing means, a first power analysis means for calculating the power and spectrum of an input signal, and a local decoding means for decoding by the local decoding means. Second power analysis means for calculating the power and spectrum of the decoded signal, and in the power calculation means, based on the results of the first power analysis means and the second power analysis means, The power loss when the bit allocation information becomes zero is calculated, and the gain adjusting means adjusts the gain for the input signal according to the power loss calculated by the power calculating means. An encoder according to claims 6 to 8. 12. A separating means for separating power information, bit allocation information, maximum value information and sample information contained in input coded data, and a partial band signal from the bit allocation information, maximum value information and sample information. Inverse quantizing means for decoding, band combining means for band combining the partial band signals decoded by the inverse quantizing means, and band combining by the band combining means according to the power information separated by the separating means A decoder comprising: a gain adjusting means for adjusting a gain of a decoded signal. 13. A separating means for separating power information, bit allocation information, maximum value information and sample information contained in input coded data, and a partial band signal from the bit allocation information, maximum value information and sample information. Dequantizing means for decoding, band synthesizing means for synthesizing the partial band signals decoded by the inverse quantizing means, and power for the partial band from the power information, bit allocation information and maximum value information from the separating section. A power loss calculating means for calculating a loss, and a maximum value adjusting means for adjusting a gain of maximum value information corresponding to the partial band separated by the separating section according to the power loss calculated by the power loss calculating means. A decoder comprising: 14. Separation means for separating bit allocation information, maximum value information and sample information included in input coded data, and inverse for decoding a partial band signal from the bit allocation information, maximum value information and sample information. Quantizing means, band synthesizing means for band synthesizing the partial band signals decoded by the inverse quantizing means, and a partial band having zero bit allocation is determined from the bit allocation information separated by the separating means. A partial band signal generating means for generating a partial band signal, a partial band signal decoded by the dequantizing means, and a partial band signal generated by the partial band signal generating means; Band combining means for synthesizing the partial band signals added by the means into a signal of the original bandwidth, In the decoder, characterized in that the bit allocation as a level of the partial band signal to be generated for the sub-band is zero, produces a noise of applying the minimum audible level. 15. Separation means for separating bit allocation information, maximum value information and sample information included in input coded data, and inverse for decoding a partial band signal from the bit allocation information, maximum value information and sample information. Quantizing means, band synthesizing means for band synthesizing the partial band signals decoded by the dequantizing means, and partial band with zero bit allocation determined from the bit allocation information separated by the separating means, and A partial band signal generating means for generating a partial band signal, a partial band signal decoded by the inverse quantization means,
And a band synthesizing unit for synthesizing the partial band signal added by the calculating unit into a signal having an original bandwidth, In the band signal generating means, the masking threshold by the partial band signal with non-zero bit allocation is obtained from the maximum value information separated by the separating means as the level of the partial band signal generated for the partial band with zero bit allocation. A decoder characterized by generating noise by applying a value level. 16. Separation means for separating bit allocation information, maximum value information and sample information included in input coded data, and inverse for decoding a partial band signal from the bit allocation information, maximum value information and sample information. Quantizing means, level information storage means for storing level information of partial band signals of past frames, band synthesizing means for band synthesizing partial band signals decoded by the dequantizing means, and separation by the separating means. A sub-band signal generating means for determining a sub-band for which bit allocation is zero from the bit allocation information, and generating the sub-band signal; and a sub-band signal decoded by the inverse quantization means,
The sub-band signal generating means includes a computing means for adding the partial band signal generated by the computing means; and a band synthesizing means for synthesizing the partial bandwidth signal added by the computing means into an original bandwidth signal. In the band signal generating means, noise is generated by applying the level stored in the level information storage means as the level of the partial band signal generated for the partial band with zero bit allocation. vessel. 17. Separating means for separating bit allocation information, maximum value information and sample information contained in input coded data, and inverse for decoding a partial band signal from the bit allocation information, maximum value information and sample information. Quantizing means, band synthesizing means for band synthesizing the partial band signals decoded by the inverse quantizing means, and a partial band having zero bit allocation is determined from the bit allocation information separated by the separating means. A partial band signal generating means for generating a partial band signal, a partial band signal decoded by the inverse quantization means,
The sub-band signal generating means includes a computing means for adding the partial band signal generated by the computing means; and a band synthesizing means for synthesizing the partial bandwidth signal added by the computing means into an original bandwidth signal. In the band signal generating means, in the partial band signal generating means, the noise that applies the maximum value information separated by the separating means as the level of the partial band signal generated for the partial band whose bit allocation is zero. A decoder for generating. 18. The sub-band signal storage means for storing a sub-band signal of a past frame output from the adding means, wherein the sub-band signal generation means outputs the sub-band signal storage means. The decoder according to claim 14, claim 15, claim 16 or claim 17, which applies a partial band signal of a past frame. 19. The sub-band signal generating means synthesizes a sub-band signal including harmonic components and sub-harmonic components of a target sub-band among the sub-band signals output from the dequantizing means. Decoder according to claims 14 to 17 for applying a signal.