JP2001356798A - Encoding and decoding device for audio signal - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an audio signal encoding and decoding device and an audio signal reproducing device which can continuously record a music signal without any break of sound recording nor the music signal. SOLUTION: A gradation encoder 22 divides an audio signal into specified blocks by a dividing filter 27 and determines a sample size of MDCT by a block size setter 29. The generation of a preecho is suppressed by making the block size of MDCT large when variation in the amplitude of a segmented sample is small and small when large.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal encoding / decoding device (hereinafter, referred to as "audio recorder").

[0002]

2. Description of the Related Art At present, a tape recorder using a magnetic tape is widely used as a small-sized audio recorder. However, tape recorders include complicated mechanical parts and electromagnetic conversion parts, so there is a limit to miniaturization, weakness to vibration, short battery life, mechanical parts wear due to repetition, random access is difficult However, there are drawbacks such as a limitation in the rising speed of recording / reproduction.

On the other hand, the progress of semiconductor technology in recent years has been remarkable, and the capacity of semiconductor memories has been significantly increased. Accordingly, various applications of the semiconductor memory to the AV field such as audio recording and image recording have been considered. Examples of applications of the semiconductor memory to audio (audio) recording are used in various products, such as answering machines, various toys, and announcement machines at stations, which have a short recording time but are still short.

[0004]

Since an audio recorder using a conventional semiconductor memory as a recording medium is configured to record an audio signal with a constant information amount, the recording time is determined by the capacity of the semiconductor memory. If the memory is exhausted during recording, it is necessary to temporarily stop recording and replace it with a new semiconductor memory, and important information will be lost or the music signal will be interrupted on the way There was a problem.

Further, since a conventional tape recorder using a magnetic tape or an audio recorder using a semiconductor memory as a recording medium is configured to record an audio signal in a fixed-length frame, audio data to be recorded is fixed. The high-efficiency encoding method that compresses to the rate is also MPEG
, Audio data of one frame (for example, 384 samples) is encoded into a fixed-length encoded frame. However, when audio data is compressed with a predetermined sound quality, the information amount of each frame is different. For example, a silent portion has little information amount, and a portion where an abrupt change such as an attack sound occurs has a large information amount. Therefore, in a coding method in which a signal of one frame is coded by a fixed-length frame, more bits are allocated to a frame having a small amount of information, and conversely, necessary bits are allocated to a frame having a large amount of information. There is a problem that it cannot be performed.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problem. Even if a semiconductor memory having a predetermined capacity is not provided with a recording time (of course, there is a recording time as a guide), the recording is performed. It is an object of the present invention to obtain a semiconductor memory audio recorder capable of continuously recording without interruption of music and without interruption of a music signal.

It is another object of the present invention to provide a semiconductor memory audio recorder capable of efficiently recording in a semiconductor memory with high sound quality (extending the recording time).

It is another object of the present invention to reproduce recorded data from a semiconductor memory recorded in variable length frames at high speed.

Further, unnecessary portions such as silent portions are skipped from the data recorded in the semiconductor memory, and only the portion in which the necessary data is recorded is reproduced, thereby achieving "early listening".
It is an object of the present invention to obtain a semiconductor memory audio recorder capable of performing the following.

[0010]

SUMMARY OF THE INVENTION An audio signal encoding / decoding device according to the present invention comprises: frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band; By dividing according to the level change based on the psychoacoustic characteristics, n-th (n is a natural number of 2 or more) first to n-th encoded data items with priorities based on the human auditory characteristics are generated. This is provided with a layering / quantizing means for performing the above.

Further, a frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and dividing the obtained conversion coefficient by using the frequency level and the quantization level as parameters to obtain a human auditory sense. Hierarchization for generating n (n is a natural number of 2 or more) first to n-th encoded data items with priorities based on characteristics
It is provided with quantization means.

Further, frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and performing hierarchization and quantization on the obtained conversion coefficient according to the following rules. It is provided with hierarchical / quantizing means for obtaining encoded data divided into n (n is a natural number of 2 or more) hierarchical levels. 1. Of the transform coefficients, the frequency band is a predetermined frequency f
A transform coefficient whose transform level is up to 1 and whose quantization level is from the MSB side to a predetermined bit number b1 is selected, and this is set as coded data S1 of the hierarchical level 1. 2. Of the transform coefficients, the frequency band is a predetermined frequency f
2 (f2 ≧ f1), and the quantization level of the transform coefficient is a predetermined number of bits b2 (b2
The transform coefficients up to ≧ b1) are selected, and the residual signal obtained by subtracting the transform coefficient of the hierarchical level 1 from this signal is used as the encoded data S2 of the hierarchical level 2. 3. Of the transform coefficients, the frequency band is a predetermined frequency f
n (fn ≧ fn−1), and
The quantization level is a predetermined number of bits bn from the MSB side.
Select conversion coefficients up to (bn ≧ bn−1).
The residual signal obtained by subtracting the transform coefficients of the hierarchical level 1 to the hierarchical level n-1 from this signal is converted into the encoded data of the hierarchical level n.
Data Sn.

Further, frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, an audible signal component based on psychoacoustic characteristics is extracted from the obtained conversion coefficient, and And a hierarchical / quantizing means for obtaining encoded data divided into n (n is a natural number equal to or greater than 2) hierarchical levels by performing hierarchical processing and quantization according to the following rules. 1. Among the transform coefficients of the audible signal component, a transform coefficient whose frequency band is up to a predetermined frequency f1 is selected, and this is used as encoded data S1 of the hierarchical level 1. 2. Among the transform coefficients of the audible signal component, a transform coefficient whose frequency band is up to a predetermined frequency f2 (f2 ≧ f1) is selected, and a residual signal obtained by subtracting the transform coefficient of the hierarchical level 1 from this signal is converted to the hierarchical level. 2 coded data S2. 3. Among the conversion coefficients of the audible signal component, a conversion coefficient whose frequency band is up to a predetermined frequency fn (fn ≧ fn−1) is selected, and further, from this signal, conversion coefficients of the hierarchical levels 1 to n−1 are determined. The subtracted residual signal is used as encoded data Sn of the hierarchical level n.

Further, division into hierarchical levels is performed in relation to level changes based on psychoacoustic characteristics.
In addition, the apparatus includes a dividing unit for dividing the input digital audio signal into blocks of a predetermined size, and the frequency converting unit converts each divided block.

Further, by dividing the frequency-converted transform coefficient according to a level change based on the psychoacoustic characteristics, it is possible to obtain hierarchically encoded audio data to which hierarchical priorities based on human auditory characteristics are given. The identification code at the hierarchical level of the hierarchically encoded audio data is input,
Decoding means for obtaining a transform coefficient corresponding to a frequency band by decoding hierarchically encoded audio data in accordance with the hierarchical level based on the identification code; and performing inverse transform on the obtained transform coefficient. Frequency inversion means for obtaining an original digital audio signal by performing the application. Further, by dividing the frequency-converted transform coefficients using the level of the frequency and the magnitude of the quantization level as parameters, hierarchically encoded audio data to which hierarchical priorities based on human auditory characteristics are given, and By inputting the identification code at the hierarchical level of the hierarchically encoded audio data as input and decoding the hierarchically encoded audio data according to the hierarchical level based on the identification code, the frequency band can be obtained. It comprises decoding means for obtaining a corresponding transform coefficient, and frequency inverse transform means for performing an inverse transform on the obtained transform coefficient to obtain an original digital audio signal.

Also, hierarchically encoded audio data that has been subjected to layering and quantization in accordance with the following rules and given hierarchical priorities based on human auditory characteristics, and the hierarchically encoded audio data: By inputting the identification code at the hierarchical level of the audio data and decoding the hierarchically encoded audio data according to the hierarchical level based on the identification code, the conversion corresponding to the frequency band is performed. It is provided with decoding means for obtaining coefficients, and frequency inverse transform means for performing an inverse transform on the obtained transform coefficients to obtain an original digital audio signal. 1. Among the conversion coefficients obtained by frequency-converting the digital audio signal, the frequency band is a conversion coefficient up to a predetermined frequency f1, and the quantization level is a conversion coefficient from the MSB side to a predetermined bit number b1. A coefficient is selected, and this is set as encoded data S1 of the hierarchical level 1. 2. Of the transform coefficients, the frequency band is a predetermined frequency f
2 (f2 ≧ f1), and the quantization level of the transform coefficient is a predetermined number of bits b2 (b2
The transform coefficients up to ≧ b1) are selected, and the residual signal obtained by subtracting the transform coefficient of the hierarchical level 1 from this signal is used as the encoded data S2 of the hierarchical level 2. 3. Of the transform coefficients, the frequency band is a predetermined frequency f
n (fn ≧ fn−1), and
The quantization level is a predetermined number of bits bn from the MSB side.
Select conversion coefficients up to (bn ≧ bn−1).
The residual signal obtained by subtracting the transform coefficients of the hierarchical level 1 to the hierarchical level n-1 from this signal is converted into the encoded data of the hierarchical level n.
Data Sn.

Also, hierarchically encoded audio data that has been subjected to layering and quantization in accordance with the following rules and given hierarchical priorities based on human auditory characteristics, and the hierarchically encoded audio data: By inputting the identification code of the hierarchical level of the audio data as input and decoding the hierarchically encoded audio data according to the hierarchical level based on the identification code, conversion corresponding to the frequency band is performed. It comprises decoding means for obtaining coefficients, and frequency inverse transform means for performing an inverse transform on the obtained transform coefficients to obtain an original digital audio signal. 1. Of the transform coefficients of the audible signal component extracted based on the human psychoacoustic characteristics, the frequency band of the transform coefficient is a predetermined frequency f
The transform coefficients up to 1 are selected, and this is used as the encoded data S1 of the hierarchical level 1. 2. From the transform coefficients of the audible signal component, a transform coefficient whose frequency band is up to a predetermined frequency f2 (f2 ≧ f1) is selected, and a residual signal obtained by subtracting a transform coefficient of the hierarchical level 1 from this signal is converted to a hierarchical level. 2 coded data S2. 3. Among the conversion coefficients of the audible signal component, a conversion coefficient whose frequency band is up to a predetermined frequency fn (fn ≧ fn−1) is selected, and further, from this signal, conversion coefficients of the hierarchical levels 1 to n−1 are determined. The subtracted residual signal is used as encoded data Sn of the hierarchical level n.

Further, frequency conversion means for converting the input digital audio signal into a conversion coefficient corresponding to a frequency band, and a conversion coefficient for converting the conversion coefficient so as to obtain a predetermined sound quality at the time of reproduction according to the characteristics of the digital audio signal. Bit allocation means for obtaining allocation information of bits to be allocated, quantization means for obtaining variable-length quantized data by quantizing a transform coefficient based on bit allocation information, quantized data and fixed-length bits Formatting means for obtaining variable-length frame data from the allocation information; frame-length detecting means for obtaining the frame length of the frame data; and writing to a recording medium according to the frame length. Address control means for controlling an address.

Further, the bit allocating means converts the energy of the transform coefficient of each band obtained by dividing the transform coefficient according to the frequency band.
, An allowable noise level detecting means for obtaining an allowable noise level of each band based on the energy of each band, and an energy and an allowable noise of each band obtained by the band dividing energy detecting means. Allocation bit detection means for obtaining the number of bits to be allocated to the transform coefficient divided into each band from the difference between the allowable noise level of each band obtained by the level detection means.

Further, a sub-band dividing means for dividing the input digital audio signal into a plurality of sub-band data, and each sub-band data according to the characteristics of the digital audio signal so as to obtain a predetermined sound quality during reproduction. Bit allocating means for obtaining bit allocation information to be allocated to the sub-band, quantizing means for quantifying each subband data based on the bit allocation information to obtain variable length quantized data, quantized data and fixed length Formatting means for obtaining variable-length frame data from the bit allocation information of
It comprises frame length detecting means for obtaining the frame length of frame data, and address control means for controlling a write address on a recording medium in accordance with the frame length.

An audio signal decoding system comprising quantized data quantized to a variable length in accordance with the characteristics of the audio signal and fixed-length bit allocation information for determining the quantization of the quantized data. An apparatus for extracting fixed-length bit allocation information from frame data, obtaining bit allocation information for obtaining the frame length of the frame data from the obtained bit allocation information, and frame length detecting means. Inverse quantum for extracting quantized data quantized to a variable length from the frame data based on the obtained frame length, and inversely quantizing the extracted quantized data based on the bit allocation information. Means. Also, frame data composed of quantized data quantized to a variable length according to the characteristics of the audio signal and fixed-length bit allocation information for determining the quantization of the quantized data.
An audio signal decoding device for extracting and decoding quantized data quantized to a variable length from data, comprising:
The fixed-length bit allocation information is extracted from the data, and the quantized data quantized to a variable length from the obtained bit allocation information.
Reproduction frame selecting means for obtaining the number of bits allocated to the data and selecting the frame data in which the number of bits exceeds a predetermined threshold value, and quantizing the selected frame data to a variable length. Inverse quantization means for extracting the quantized data thus obtained and inversely quantizing the extracted quantized data based on the bit allocation information.

[0022]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. In the first embodiment, for the sake of simplicity, the number of hierarchical code block divisions for audio encoding will be described as 2. However, the basic concept is the same even when the number of divisions increases. In FIG. 1, reference numeral 1 denotes an input terminal of an audio signal, 2 denotes an audio amplifier for adjusting a required audio level in the next stage, 3 denotes an A / D converter for converting an audio signal into a digital signal, and 4 denotes a hierarchical code of the digital audio signal. 5, a semiconductor memory that is a recording medium for audio signals, 6 writes an audio signal from the hierarchical encoder 4 to a predetermined address in the semiconductor memory 5, and
A memory address controller for reading an audio signal from a predetermined address and sending it to a hierarchical decoder 7; a hierarchical decoder 7 for decoding the audio signal encoded by the hierarchical encoder 4; A D / A converter 9 for converting to an audio signal, 9 is an audio amplifier for adjusting the output of the D / A converter to a required audio level in the next stage, 10 is an audio signal output terminal, 1
4 is a clock generator.

FIG. 2 and FIG. 3 show that the hierarchical encoder 4
FIG. 3 is a diagram illustrating a configuration example in which a digitized audio signal is subjected to two-division encoding. In the configuration example of FIG.
In the digital audio signal of bits, the upper 8 bits are divided into a hierarchical code block 1 and the lower 8 bits are divided into a hierarchical code block 2.

In the configuration example shown in FIG. 3, the audio frequency is divided into four sub-bands by a sub-band dividing filter 15 for four divisions, and a hierarchical code block 1 and a hierarchical code Bit allocation to block 2 is determined, and control is performed so that the total signal amount of hierarchical code block 1 and hierarchical code block 2 for each subband is divided into two.

In each of the configuration examples shown in FIGS. 2 and 3, by assigning the higher-order bit side of the quantized bits to the higher-order hierarchical code block 1, even if the lower-order hierarchical code block 2 is lost. Although the sound quality is deteriorated, the audio signal can be reproduced.

Further, in the band division method shown in FIG. 3, by optimizing the bit allocation in consideration of the human audibility, even if the audio signal is only the hierarchical code block 1, the sound quality is deteriorated. Can be reduced as much as possible.

FIG. 4 shows a memory map on the semiconductor memory 5, which comprises a control data area for recording an identification code of an audio hierarchical code block, an audio area A for recording an audio signal, and an audio area B. FIG. 5 shows how audio signals are recorded in the semiconductor memory 5 over time.

Next, the operation of the recording system will be described first. In FIG. 1, an audio signal input to an input terminal 1 is amplified to a predetermined level by an audio amplifier 2, converted to a digital signal by an A / D converter 3, and input to a hierarchical encoder 4. In the hierarchical encoder 4, a signal for recording the digitized audio signal in the hierarchical code block 1 by the method shown in FIG.
The signal to be recorded in the hierarchical code block 2 is divided into two.

The coded signal divided into two is recorded in the semiconductor memory 5 as shown in a memory map of FIG. At the start of recording, as shown in FIG. 5A, at the start of recording, the hierarchical code block 1 encoded by the encoder 4 is stored in the audio area A, and the hierarchical code block 2 is stored in the audio area B. Recorded in order. FIG. 5B shows a state in which audio signals are sequentially recorded and the memory 5 is almost full.

FIG. 5 (c) shows an example of overwriting an audio signal writing area (in this case, audio area B) of the lower priority hierarchical code block 2 in order to record a further continuous audio signal. , Only the hierarchical code block 1 of the audio signal is recorded.

FIG. 5D shows a state where the overwriting is performed and the recording area is full. The memory address controller 6 controls the memory address in accordance with the signal from the memory capacity detector 13 so that the flow described with reference to FIG. 5A and 5B, the identification code at the hierarchical level is “hierarchical level 1”, and the hierarchical code block “00” is used.
In FIGS. 5C and 5D, the hierarchical code block “01” is recorded as “hierarchical level 2”.

At the time of reproduction, first, a hierarchy level identification code is checked by the hierarchy level identification code reproducing unit 12. B
, And outputs the reproduced audio signal to the hierarchical decoder 7. When the identification code of the hierarchy level is “01”, that is, “hierarchy level 2”, the audio area A of the semiconductor memory 5 is accessed sequentially (the audio area B is accessed after the audio area A ends). ), And outputs the reproduced audio signal to the hierarchical decoder 7.

The hierarchical decoder 7 is configured to decode the reproduced signal from the semiconductor memory 5 into a digital audio signal in any of "hierarchical level 1" and "hierarchical level 2"("hierarchicallevel"). In the case of the reproduction signal of "2", zero may be input to the signal of the hierarchical code block 2 by the hierarchical level identification code code).

The output of the hierarchical decoder 7 is a D / A converter 8
Is converted into an analog audio signal, amplified by the audio amplifier 9 to a predetermined audio level, and output from the output terminal 10. Note that a clock necessary for the entire system at the time of recording and reproduction, such as a sampling clock, is supplied from the clock generator 14.

Embodiment 2 FIG. 6 shows a block circuit according to a second embodiment of the present invention, in which the same reference numerals as in FIG. 1 denote the same parts, and reference numeral 17 denotes a recording capable of recording all of the audio signals encoded by the hierarchical encoder 4. A first semiconductor memory having a sufficient capacity, 18 is a second semiconductor memory having a predetermined capacity for storing audio data (such as a removable memory card), and 19 is a hierarchical level converter, which is a first semiconductor memory. From the memory 17 to the second semiconductor memory 1
When dubbing the audio signal to the memory 8, the memory capacity of the second semiconductor memory 18 is justified according to the audio recording time output from the memory capacity detector 13.
The entire signal amount is controlled by omitting layer code blocks having lower priority in block coding.

Reference numeral 20 denotes a first memory address controller, which controls the audio signal from the hierarchical encoder 4 to be recorded at a predetermined address in the first semiconductor memory 17;
When dubbing from the first semiconductor memory 17 to the second semiconductor memory 18, the transfer speed is increased and control for reading data from the first semiconductor memory 17 is performed. Reference numeral 21 denotes a second memory address controller which controls reading of an audio signal from a second semiconductor memory from a predetermined address;
When dubbing from the first semiconductor memory 17 to the second semiconductor memory 18, the transfer speed is increased to control writing data to the second semiconductor memory.

Next, the operation of the recording system will be described. The audio signal input to the input terminal 1 is amplified to a predetermined level by the audio amplifier 2, converted to a digital signal by the A / D converter 3, and input to the hierarchical encoder 4. In the hierarchical encoder 4, as shown in FIG. 2 or FIG. 3, the signals are encoded into signals of two hierarchical code blocks having priority. Here, as in the first embodiment, even when a hierarchical code block with a low priority is lost and only a hierarchical code block with a high priority is left, the audio signal can be normally reproduced though the sound quality is deteriorated, as in the first embodiment.
The audio signal divided into two blocks and encoded is recorded in the first semiconductor memory 17 according to the address signal from the first memory address controller 20. Since the first semiconductor memory 17 has a sufficient capacity, all the input audio signals are recorded.

Next, the dubbing operation from the first semiconductor memory 17 to the second semiconductor memory 18 will be described.
The memory capacity detector 13 detects the recording time of the audio signal written in the first semiconductor memory 17 and sends this detection signal to the hierarchical level converter 19, where the memory capacity of the second semiconductor memory 18 is determined. A comparison is made to determine whether or not the signal needs to be reduced, and if so, how much the signal should be reduced. If so, the hierarchical level is changed according to the ratio. That is, a hierarchical code block having a low priority is deleted.

In the second embodiment, the number of data divisions is set to 2. However, as the number of divisions increases, the recording time and the degree of deterioration in sound quality can be more finely controlled. It can be used efficiently (the recording time and the degree of deterioration of sound quality have a linear relationship).

Further, this hierarchy level is sent to the hierarchy level identification code generator 11, and the identification code is simultaneously recorded in the control area of the second semiconductor memory 18. At the time of dubbing, the addressing of the first memory address controller 20 and the second memory address controller 21 is operated at high speed, and high-speed dubbing is performed within a range allowed by the memory access speed. This high-speed dubbing function is an important function to make it appear as if recording is performed by one recording medium even though there are two recording media (semiconductor memories).

At the time of reproduction, first, the hierarchy level identification code reproducer 12 determines the identification code from the control area of the second semiconductor memory 18 and sends the result to the second memory address controller 21. Second memory address controller 21
Are the second addresses in order from a predetermined address according to the hierarchy level.
The readout signal is read from the semiconductor memory 18 of FIG.

The hierarchical decoder 7 decodes a reproduced signal of any hierarchical level in accordance with each signal. The output of the hierarchical decoder 7 is converted into an analog audio signal by a D / A converter 8, amplified to a predetermined audio level by an audio amplifier 9, and output from an output terminal 10. Note that a clock such as a sampling clock required for reproduction is supplied from the clock generator 14.

Embodiment 3 Hereinafter, a third embodiment of the present invention will be described with reference to FIGS. In the third embodiment, a description will be given assuming that the number of divisions of the hierarchical code block in the hierarchical coding is 4, but the concept is the same even when the division number is different. In FIG. 7, the same reference numerals as those in FIG. 1 denote the same parts. Reference numeral 22 denotes a hierarchical encoder for digital audio signals, and reference numeral 23 denotes a memory address controller. Is written to a predetermined address for each layer, and an audio signal of each layer is read from the predetermined address. Further, the capacity of the semiconductor memory 5 is detected, and when the semiconductor memory 5 is full, the data is already written. Among the data, control is performed so as to overwrite temporally continuous audio signals from the memory area of the data written as the hierarchical code block having the lower priority by the hierarchical encoder 22. Reference numeral 24 denotes a hierarchical decoder that decodes the audio signal encoded by the hierarchical encoder 22.

FIGS. 8 and 9 show the hierarchical encoder 22.
FIG. 4 is a diagram for describing the content of performing hierarchical coding by dividing a digital audio signal into four parts. First, in FIG.
The signals are classified based on the input original audio signal. That is, the signals are roughly classified into three types: a signal that cannot be heard originally due to the minimum audibility, which is a human auditory characteristic, a signal that cannot be heard due to the masking effect, and a signal that can be heard further. Next, only signals that can be heard from these are selected, and are further divided into four hierarchical levels according to the frequency characteristics shown in FIG.

FIG. 10 shows the hierarchical encoder 22 of the hierarchical level 4.
Is a diagram for explaining the configuration of the audio signal and the recording method for the semiconductor memory 5, wherein 27 is a division filter for dividing the audio signal converted into a digital signal by the A / D converter 3 into predetermined blocks, and 28 is a division filter. MD signal
MDCT converter for performing orthogonal transformation by CT, 29 is a block size setting unit for setting a conversion block size of MDCT 28 according to a change in an input signal, and 30 is a grouping device for grouping coefficients of MDCT 28 according to a critical band based on psychological psychology. , 31 are FIGS. 8, 9
According to the classification of the audio signal shown in the above, the inaudible signal is removed, and the audible signal is hierarchized into four hierarchical levels and quantized. A dynamic bit allocator 33 for deciding whether to allocate bits, a scale factor calculator 33 for determining a scale factor in accordance with the blocks of the MDCT 28, and a formatting device 34 for formatting a recording for recording in the semiconductor memory 5. On the right side of FIG. 10, the concept of recording hierarchically encoded data in the semiconductor memory 5 is shown.

FIG. 11 is a diagram showing a memory map on the semiconductor memory 5, which is composed of a control data area for recording a hierarchy level identification code and an audio area for recording an audio signal (hierarchical level 1 to hierarchical level 4). In addition, (1) to (4) in FIG. 11 show the recording state of the audio signal in the semiconductor memory 5 as time elapses. That is, (1) shows a state in which an audio signal coded at the hierarchical level 4 is recorded until the memory is full, and (2) shows a case in which the audio signal is recorded continuously after the situation of (1). In the state where the recording area of level 4 is overwritten with the hierarchical levels 1 to 3, the state (3) is the case where the situation of (2) is passed and the audio signal is further continuously recorded. (4) is a state in which the audio signal has been continuously recorded after the situation of (3), and the recording area of the hierarchical level 2 has been overwritten at the hierarchical level 1 (on the memory). Are all at the hierarchical level 1).

Next, the operation of a portion different from that of the first embodiment during recording will be described. The audio signal converted into a digital signal by the A / D converter 3 and input to the hierarchical encoder 22 is output from the hierarchical encoder 22 as shown in FIGS.
Is converted to hierarchically encoded data and recorded in the semiconductor memory 5 by the method shown in FIG. In this hierarchical encoder 22,
First, the audio signal is divided into predetermined blocks by the division filter 27, and a sample size for performing MDCT is determined by the block size setting unit 29. If the change in the amplitude of the extracted sample is small, the block size of the MDCT is increased, and if the change in the amplitude is large, the MDCT block size is reduced to suppress the occurrence of pre-echo.

In the MDCT converter 28, the divided filter 2
Performs MDCT conversion for each frequency band divided by 7,
Each transform coefficient is grouped for each critical band by a grouping unit 30 based on psychoacoustics, and divided into four bands by a hierarchical / quantizer 31 according to the following rules. FIG. 9 shows the band and range.

1. First, as the hierarchical level 1, a coefficient from a low-frequency coefficient to a predetermined frequency band f1 is selected from among the coefficients divided into a plurality of frequency bands, and the quantization level of the coefficient is also predetermined from the MSB side. The number of bits is selected, and this is set as encoded data S1 of the hierarchical level 1.

2. Next, as the hierarchy level 2, a coefficient up to a predetermined frequency band f2 higher than f1 is selected, and the quantization level of the coefficient is also selected from the MSB side to a predetermined bit number larger than the hierarchy level 1, The residual signal obtained by subtracting the signal component of the coefficient of the hierarchical level 1 corresponding to each coefficient from the above is defined as the encoded data S2 of the hierarchical level 2.

3. Next, as the hierarchy level 3, a coefficient up to a predetermined frequency band f3 higher than f2 is selected, and the quantization level of the coefficient is also selected from the MSB side to a predetermined number of bits larger than the hierarchy level 2, and this signal Is obtained by subtracting the signal components of the coefficients of the hierarchical levels 1, 2, and 3 corresponding to each coefficient from the coded data S3 of the hierarchical level 3.
And

4. Similarly, as hierarchical level 4, a coefficient up to a predetermined frequency band f4 higher than f3 is selected,
Also, the quantization level of the coefficient is selected from the MSB side as a predetermined number of bits larger than that selected at the hierarchical level 3, and from this signal, the hierarchical levels 1, 2 and 3 corresponding to each coefficient are selected.
The residual signal obtained by subtracting the signal component of the coefficient is used as the encoded data S4 of the hierarchical level 4.

On the other hand, the dynamic bit allocation unit 32 determines how bits are allocated to each frequency band in units of MDCT blocks, and the scale factor calculator 33 determines the maximum value of each coefficient in units of MDCT blocks. Extract the value and normalize the value of each sample with this maximum. In the formatting device 34, the scale factor, bit allocation, and hierarchical data (S1, S2, S
3, and S4) are formatted and sent to the semiconductor memory 5.

The hierarchized audio signal sent to the semiconductor memory 5 is sent to the memory address controller 23 in FIG.
As shown on the right side of FIG. 0, first, the hierarchical level 4 is classified and recorded in the semiconductor memory 5 for each hierarchical level. When the semiconductor memory 5 becomes full and the recording memory is exhausted, a continuous audio signal is overwritten at the hierarchical levels 1 to 3 in the memory area written at the hierarchical level 4 next. FIG. 11 shows this state on a memory map. The above description shows the recording state of FIG. 11 (1) → (2). Hierarchical level identification code generator 25
Generates an identification code "11" in the state of FIG. 11 (1) for the maximum hierarchy level 4 and an identification code "10" in the state of FIG. It is recorded in the control data area of the memory 5.

Similarly, at the same time as the recording state proceeds as shown in FIG. 11 (2) → (3) → (4), the hierarchical level also changes from 3 → 2 → 1, and the identification code to be recorded is “10” → “01”. → "0
0 ". In this case, the recording time is the maximum when the audio signal of the hierarchy level 1 is finally recorded on the semiconductor memory 5 and the information amount of each hierarchy level (1 to 4) is the same. Then, the recorded audio quality is lower than that in the state of FIG. 11A, but the recording time is quadrupled.

At the time of reproduction, first, the hierarchy level identification code is checked by the hierarchy level identification code reproducer 26. This information is received by the memory address controller 23, and FIG.
The audio signal recorded according to the memory map 1 is read from the semiconductor memory 5. First, in the case where the identification code is “11”, that is, at the hierarchical level 4, FIG.
The signals of hierarchy levels 1 to 4 are read out according to the memory map of (1). If the identification code is "10", that is, level 3, signals of hierarchical levels 1 to 3 are read out according to the memory map of FIG. Similarly,
11 when the identification code is “01” and the hierarchical level is 2
According to the memory map of (3), signals of hierarchy levels 1 and 2 are identified by the identification code "00" and the signal of hierarchy level 1 is used in FIG.
The signal of the hierarchy level 1 is read out according to the memory map of 1 (4).

In the hierarchical decoder 24, the signal read from the semiconductor memory 5 is converted to the hierarchical level identification code reproducer 2.
6, the decoding is performed in accordance with the hierarchical level. The output of the hierarchical decoder 24 is converted into an analog audio signal by the D / A converter 8, and is output from the audio output terminal 10 via the audio amplifier 9 for amplifying to a predetermined audio level. Clocks necessary for the entire system such as a sampling clock at the time of recording / reproduction are clock generator 1
4.

Embodiment 4 Next, a fourth embodiment of the present invention will be described with reference to FIGS. FIG. 12 is a block circuit diagram of the fourth embodiment. The same reference numerals as those in FIG. 7 denote the same parts. In FIG. 12, an address switch 35 switches between a memory address from a write address generator 36 and a memory address from a read address generator 37. Write address generator 3
Reference numeral 6 denotes an address for writing a hierarchically encoded audio signal in the semiconductor memory 5. The read address generator 37 generates an address for reading the hierarchically encoded audio signal from the semiconductor memory 5 according to the reproduction clock from the clock frequency divider 39. Numeral 38 denotes a hierarchical level determiner, which changes the decoding operation time given to the hierarchical decoder 24 depending on the reproduction speed of the audio signal (that is, if the reproduction speed increases, the operation time given for decoding a predetermined number of audio samples becomes The shorter the playback speed, and conversely the slower the playback speed, the longer the computation time given to decoding a given number of audio samples:
Hierarchical encoders and decoders are DSP (Digital S)
In many cases, software processing is performed by using a signal processor, etc., and the DSP itself performs a considerably high-speed calculation. Therefore, even if the reproduction speed of the audio signal is increased, the calculation time of the DSP is similarly reduced. High speed cannot be achieved), and a hierarchical level that can be decoded within a given operation time is determined and the hierarchical decoder 2
Inform 4 Reference numeral 39 denotes a clock frequency divider which divides the frequency of the clock generated by the clock generator 14 based on a signal from the reproduction speed setting switch and sends it to the read address generator 37. Reference numeral 40 denotes a switch constituting a reproduction speed setting device.

FIG. 13 is a diagram showing the relationship between the audio decoding time and the reproduction speed according to the fourth embodiment.
(B) shows a case where all four hierarchical level signals can be decoded during normal reproduction. On the other hand, FIG.
(D) shows a case in which the reproduction speed is doubled, and only two hierarchical levels of signals can be decoded in time.

Next, the operation of a portion different from that of the third embodiment during recording will be described. The audio signal hierarchically encoded by the hierarchical encoder 22 is output to the write address generator 3.
The write address generated in step 6 is the address switch 3
5 is given to the semiconductor memory 5 and recorded at a predetermined address.

At the time of reproduction, first, the address switch 3
5, the read address of the semiconductor memory 5 is switched to the read address generator 37. In the read address generator 37, the reproduction speed is set by the reproduction speed setting unit 40 (in this example, UP and DOW are set for the normal reproduction speed).
The clock divider 39 generates a clock to be supplied to the read address generator 37 in accordance with the read speed. An audio signal is read from the semiconductor memory 5 according to the set reproduction speed, and is decoded by the hierarchical decoder 24. In this case, since the decoding operation time given to the layer decoder 24 changes according to the reproduction speed of the audio signal as shown in FIG. 13, the decoding hierarchy level that can be decoded within the given operation time is changed. The determination is made by the hierarchy level determiner 38.

In the normal reproduction shown in FIGS. 13A and 13B,
All four hierarchical level signals can be decoded, while FIG.
When the reproduction speed in (c) and (d) is doubled, it indicates that decoding operation can be normally performed up to signals of two hierarchical levels. When the reproduction speed is doubled, the reproduced audio signals all have a double frequency, that is, 5k.
Hz signal component shifts to 10 kHz and 10 kHz signal component shifts to 20 kHz.
The signal above Hz will exceed the audible band,
Decoding of hierarchical level 3 or 4 having many high-frequency components is almost unnecessary. From this, as shown in FIGS. 13C and 13D, a method of decoding only the hierarchical levels 1 and 2 is a very suitable method. The output of the hierarchical decoder 24 is converted into an analog audio signal by the D / A converter 8, and is output from the audio output terminal 10 via the audio amplifier 9 for amplifying to a predetermined audio level.

Embodiment 5 Hereinafter, a fifth embodiment of the present invention will be described with reference to the drawings. The block circuit diagram of the fifth embodiment is the same as that of the third embodiment shown in FIG. Also, the conceptual diagram of the hierarchical coding of the semiconductor memory audio recorder according to the fifth embodiment is the same as that of FIG.

FIG. 14 is a diagram showing a manner of dividing the hierarchical level of the hierarchical encoder 22 according to the fifth embodiment. The hierarchical encoding method is different from that of the third embodiment. Hereinafter, the code block division number of the hierarchical coding is set to 4 similarly to the third embodiment, and the classification of the input original audio signal is the minimum which is the human auditory characteristic similarly to FIG. 8 of the third embodiment. Signals that cannot be heard originally due to audibility, signals that cannot be heard due to the masking effect, and signals that can be heard are roughly classified into three, and only signals that can be heard are selected from these signals. Divide into levels.

That is, in FIG. 14, the audible component exceeding the masking level is divided into four in the frequency direction from the hierarchical level 1 to the hierarchical level 4 so that the information amount is equal, and the entire information amount is equal to the desired bit rate. To satisfy.

FIG. 15 shows the hierarchical encoder 2 according to the fifth embodiment.
FIG. 2 is a diagram showing a configuration 2 and a method of recording on a semiconductor memory 5. In the figure, reference numeral 41 denotes a subband n-division filter.
The audio signal converted into a digital signal by the A / D converter 3 is divided into a plurality of sub-bands. Reference numeral 42 denotes an audible component extracting means for converting an audio signal into a frequency domain by FFT transform, and extracting only audible components by masking based on auditory characteristics. Reference numeral 43 denotes an information amount calculating unit for each subband of each frame. The audible component extracted by the audible component extracting unit 42 is 1 bit per 6 dB.
The information amount of each subband is calculated by allocating the information amount of. Reference numeral 44 denotes an information amount calculating unit for each frame, which calculates the information amount per encoded frame by summing the information amounts of the respective subbands obtained by the information amount calculating unit 43 for each subband.

Reference numeral 45 denotes an information amount per frame setting unit for setting an information amount C per frame based on a desired encoding rate. Reference numeral 46 denotes an information amount control circuit. According to the calculated information amount of each frame and the average allocated information amount which is the average value of the information amounts allocated to the frames encoded up to that point, the final average information amount is information per frame based on the coding rate. The allocation information amount to be allocated to each frame is calculated so as to match the setting information amount C obtained by the amount setting unit 45. 47 is a band determining means of each layer,
Based on the information amount for each sub-band of each frame calculated by the information amount calculation means 43 and the allocation information amount calculated by the information amount control circuit 46, the optimum hierarchical level (K1 to K4) for giving the allocation information amount Determine the coding band. Reference numeral 48 denotes a bit allocation circuit, which re-allocates bits to audible components in an allocated band for each layer according to the size according to band information allocated to each layer obtained by the band determining means 47 for each layer. The quantization circuit quantizes each sub-band data according to the bit allocation information obtained by the bit allocation circuit 48.

Numeral 50 denotes a hierarchical coding formatter, which allocates band information of each layer obtained by the band determining means 47 of each layer, bit allocation information obtained by the bit allocation circuit 48, and obtains information from the quantizer 49. The obtained data is hierarchically encoded and formatted. FIG.
5 shows the concept of recording hierarchically encoded data in the semiconductor memory 5.

The memory map on the semiconductor memory 5 and the procedure for writing the audio signal are the same as those in FIG.

Next, a description will be given of an operation at the time of recording which is different from the third embodiment. In the hierarchical encoder 22, the audio signal is converted into hierarchically encoded data by the method shown in FIGS. 8 and 14, and is recorded in the semiconductor memory 5.
In the hierarchical encoder 22, first, the audio signal is divided into n subbands by a subband n-division filter 41, and at the same time, the audio signal is orthogonally transformed into a frequency domain by FFT transform in an audible component extracting means 42. A masking level based on auditory characteristics is determined in the region, and audible components are extracted. As shown in FIG. 16A, the difference between the frequency spectrum and the masking level is the audible component. Further, the audible components are collected for each sub-band by sub-band division, and the audible components for each sub-band are extracted as shown in FIG.

Next, the information amount calculating means 43 gives an information amount of 1 bit to the audible component 6 dB, thereby giving an information amount of 1 bit.
As shown in FIG. 6 (c), the information amount for each sub-band is calculated. Then, the information amount calculating means 4 for each frame
In step 4, the information amounts of the n subbands are added, and the information amount of one frame is calculated. The information amount per frame setting unit 45 based on the coding rate calculates the information amount C per frame based on the bit rate by setting a desired coding rate, and sends it to the information amount control circuit 46.

In the information amount control circuit 46, the final average allocated information amount is determined by the information amount setting unit per frame based on the coding rate in accordance with the average information amount per frame allocated to the frames encoded up to that time. Is determined for the information amount calculated by the information amount calculation means 44 of each frame so as to match the information amount C set in the above. For example, let s be the total number of bits allocated to frames encoded so far.
If um and the number of encoded frames are count, sum
Is divided by count to obtain the average allocated information amount (M bar).
Is larger than the information amount C per frame converted from the coding bit rate, the information amount control coefficient (K = C / (M bar)
Is multiplied by the information amount (m) calculated by the information amount calculating means 44 of each frame to calculate an allocated information amount (M = mK).

Next, the band determining means 47 for each layer uses the information amount for each subband obtained by the information amount calculating means 43 and the allocated information amount obtained by the information amount control circuit 46 for each layer. A band that can be covered by the allocation information amount is calculated, and is set as a band of each layer. For example, if the value obtained by dividing the audio signal into 16 subbands and calculating the information amount for each subband is as shown in FIG. 17, the allocation information amount for that frame is 4
In the case of 3 bits, the amount of information allocated to each hierarchical level from the hierarchical level 1 (K1) to the hierarchical level 4 (K4) is made equal, so that the information amount that can be allocated to each layer is It becomes 10 bits.

The allocation of the information amount to each frame is performed in the following procedure. First, an assigned band of K1 is obtained. Since the information amount of the sub-band 1 of the lowest band is 10 bits, the information amount of K1 is 10 bits only with this sub-band. Therefore, the band that can be allocated as the encoding band of K1 is 1 sub-band. . Next, an assigned band of K2 is obtained. When the information amount of subband 2 and subband 3 is added from the subbands equal to or greater than the band allocated to K1, 1
The number of bits becomes 0 (6 + 4), which coincides with 10 bits assigned to K2. Therefore, the bands that can be allocated as the K2 coding band are a few subbands. Similarly, when the information amount from subband 4 to subband 6 which is a subband equal to or larger than the band allocated to K2 is added, 1
It becomes 0 bits, which is equal to the allocation information amount of K3, and when the information amount from subband 7 to subband 10 is added, 10 bits
Bits, which is equal to the allocation information amount of K4.

As a result, the bands that can be allocated to each layer are, as shown in FIG. 17, one subband, two subbands, four to six subbands, and seven to ten subbands. In the bit allocation circuit 48,
Based on the allocated band information to each layer obtained by the band determining means 47 of each layer and the audible component obtained by the audible component extracting means 42, the audible component in the allocated band of each layer is
Re-bit allocation is performed according to the size. For example,
As shown in FIG. 17, when bands are allocated to the respective hierarchical levels K1 to K4, the K1 band is one subband of only subband 1, the K2 band is two subbands of subbands 2 and 3, and the K3 band is a subband. 3 subbands of bands 4 to 6, K4
The bands are four subbands of subbands 7 to 10.

Therefore, at the hierarchical level 1, one bit per 6 dB is allocated to the audible component of the subband 1, and then at the hierarchical level 2, the size of the audible component is According to 1 per 6dB
The bits are re-bit-assigned, and the hierarchy level 3 is re-assigned to the audible components of sub-bands 4, 5, and 6, and the hierarchy level 4 is re-assigned to the audible components of sub-bands 7, 8, 9, and 10. The quantizer 49 quantizes the sub-band data obtained from the sub-band n-division filter 41 according to the bit allocation information for each sub-band obtained from the bit allocation circuit 48. In the hierarchical coding formatter 50, the band determining means 47 of each layer
The bit allocation information and the quantized data are formatted for each layer according to the obtained allocation band information of each layer, and sent to the semiconductor memory 5. In the data encoded by the hierarchical encoder 22 as described above, the information amount of each frame is variable length, but the information amount of each layer included in one frame is the same information amount.

The hierarchized audio signal sent to the semiconductor memory 5 is supplied to the memory address controller 23 as shown in FIG.
As shown on the right side of 5, the data is first classified and recorded in the semiconductor memory 5 at the hierarchical level 4. When the semiconductor memory 5 becomes full, a continuous audio signal is overwritten at the hierarchy level 3 in the memory area where the hierarchy 4 has been written. Although FIG. 15 shows the concept and is written so that the information amount of each frame is equal, the semiconductor memory 5 is actually encoded in a variable-length frame.
It is controlled so that it can be stored according to the information amount of each frame.

FIG. 11 shows this state on the memory map. The above description shows the recording state shown in FIG. 11 (1) → (2). In the hierarchical level identification code generator 25, in the state of FIG.
The identification code "10" is recorded because the state of (2) is the maximum hierarchy level 3. Similarly, FIG. 11 (2) → (3)
→ (4) As the recording state progresses, the hierarchy level is also 3 → 2
→ 1 and the identification code to be recorded is also “10” → “01”
→ "00". In this case, the recording time is the maximum when the audio signal of the hierarchy 1 is finally recorded on the semiconductor memory. If the information amount of each hierarchy (1 to 4) is the same, the state of FIG. Although the recorded audio quality deteriorates, the recording time is quadrupled.

At the time of reproduction, first, the hierarchy level identification code is checked by the hierarchy level identification code reproducer 26. This information is received by the memory address controller 23, and FIG.
The audio signal recorded according to the memory map 1 is read from the semiconductor memory 5. First, when the identification code is "11", that is, at the hierarchical level 4, FIG.
In accordance with the memory map shown in FIG. When the identification code is "10", that is, at the level 3, the hierarchical level 1 according to the memory map of FIG.
3 signals are read out.

Similarly, when the identification code is "01" and the hierarchical level is 2, the signals of the hierarchical levels 1 and 2 are converted to the hierarchical level 1 by the identification code "00" according to the memory map of FIG. In this case, the signal of the hierarchical level 1 is read out according to the memory map of FIG. The hierarchical decoder 24 is configured to decode the signal read from the semiconductor memory 5 according to the identification level from the hierarchical level identification code reproducer 26 according to the hierarchical level. The output of the hierarchical decoder 24 is the D / A converter 8
The audio signal is converted into an analog audio signal, and is output from an audio output terminal 10 through an audio amplifier 9 that amplifies the audio signal to a predetermined audio level.

Embodiment 6 FIG. Next, a sixth embodiment of the present invention will be described. FIG. 18 is a diagram showing the configuration of the hierarchical encoder of the semiconductor memory audio recorder according to the sixth embodiment. In the figure, reference numeral 41 denotes a sub-band n-division filter, which divides an audio signal converted into a digital signal by the A / D converter 3 into a plurality of sub-bands. 4
Reference numeral 2 denotes an audible component extracting unit which converts an audio signal into a frequency domain by FFT transform, and extracts only audible components by masking based on auditory characteristics. Reference numeral 45 denotes an information amount setting unit per frame based on a coding rate, and sets an information amount C per frame based on a desired coding rate. Reference numeral 48 denotes a bit allocation circuit, based on the band information to be assigned to each layer in one cycle obtained by the band determining means 54 for each layer, and for each frame, an audible component in an assigned band for each layer according to its size. Re-bit allocation is performed. A quantization circuit 49 quantizes each sub-band data according to the bit allocation information obtained by the bit allocation circuit 48.

Numeral 50 denotes a hierarchical coding formatter, which is obtained by the band determiner 54 of each layer, the bit allocation information obtained by the bit allocation circuit 48, and the bit allocation information obtained by the quantizer 49. Data is hierarchically encoded and formatted. Reference numeral 51 denotes an average information amount calculation unit which takes an average value of the audible components extracted by the audible component extraction unit 42 per fixed time and assigns an information amount of 1 bit per 6 dB to the average value to thereby obtain an average information amount of each frame. The average information amount for each sub-band is calculated. Reference numeral 52 denotes an instantaneous information amount calculating unit for each frame, which calculates an information amount for each encoded frame from the maximum subband information SBmax having an audible component obtained by the audible component extracting unit 42. Reference numeral 53 denotes an information amount control circuit, which determines the final average information amount according to the instantaneous information amount obtained by the instantaneous information amount calculation means 52 of each frame and the average allocated information amount allocated to the frames encoded up to that time. In order to match the setting information amount C obtained by the information amount setting unit 45 per frame based on the coding rate,
A section allocation information amount for one cycle is calculated. 54 is a band determining means of each layer, and the average information amount for each subband of each frame calculated by the average information amount calculating means 51;
Based on the section allocation information amount calculated by the information amount control circuit 53, a coding band of each layer in one cycle of each layer level (K1 to K4) optimal for providing the allocation information amount is determined.

FIG. 19 is a flowchart showing the operation of calculating the average information amount for each subband by the average information amount calculating means according to the sixth embodiment.

FIG. 20 is a flowchart showing a section allocation information amount calculation operation in the information amount control means according to the sixth embodiment.

FIG. 21 is a flowchart showing the operation of determining the allocated band in the band determining means 54 of each layer according to the sixth embodiment.

FIG. 22 is a flow chart showing a bit allocation operation in bit allocation circuit 48 according to the sixth embodiment.

Next, the operation will be described. In the hierarchical encoder 22, first, the audio signal is divided into n subbands by the subband n division filter 41, and at the same time, the audio signal is
The orthogonal transform into the frequency domain is performed by the FT transform, a masking level based on the auditory characteristics is obtained in the frequency domain, and an audible component is extracted. As shown in FIG. 16A, the difference between the frequency spectrum and the masking level is the audible component.
Further, the audible components are collected for each sub-band by sub-band division, and the audible components for each sub-band are extracted as shown in FIG.

Next, in the average information amount calculating means 51, FIG.
As shown in the flowchart of FIG. 9, the audible component for each sub-band extracted by the audible component
The average value for the X frames of the audible component is calculated by accumulating and adding the frames, and dividing by X after the addition, and by giving the information amount of 1 bit per 6 dB to the average audible component, the average information for the X frame is obtained. The amount is calculated and used for the band determination in the next one cycle (for X frames). The instantaneous information amount calculating means 52 for each frame estimates the instantaneous information amount from the maximum band (hereinafter, referred to as SBmax) in which the audible component is obtained by the audible component extracting means 42 by the following method.

FIGS. 23 to 25 show the maximum band SB in which the audible component exists in the instantaneous information amount determining means according to the sixth embodiment.
FIG. 5 is a diagram showing the relationship between max and the amount of information. As can be seen from these data, since SBmax and the amount of information are substantially proportional, the amount of information can be estimated from SBmax. In the sixth embodiment, the instantaneous information amount is obtained by estimation using the relationship between SBmax and the information amount. However, an information amount of 1 bit per 6 dB is given to the audible component, and the information amount of each subband is added. It may be calculated.

The information amount per frame setting unit 45 based on the coding rate calculates the information amount C per frame based on the bit rate by setting a desired coding rate. Can be In the information amount control circuit 53, the final average allocated information amount is set by the information amount setting unit per frame based on the coding rate according to the average information amount per frame allocated to the frames encoded up to that time. The one-cycle section allocation information amount is determined for the instantaneous information amount calculated by the instantaneous information amount calculation means 52 of each frame so as to match the information amount C thus obtained.

The operation of the information amount control circuit 53 is shown in FIG.
0 is shown in the flowchart. Section allocation information amount MN bar
Is calculated according to the following equation, for example, according to the average allocation information amount M bar of frames that have been encoded for one cycle of the X frame and for each cycle. MN bar = C
+ (CM bar) where C is the information amount per frame based on the coding rate, and the MN bar is controlled so that the average allocated information amount finally becomes this value.
Also, during one cycle, the instantaneous allocation information amount M is calculated by multiplying the instantaneous information amount m calculated by the instantaneous information amount calculation circuit 52 for each frame by the information amount control coefficient K. The average allocation information amount M bar and the information amount control coefficient K based on the average value of the instantaneous allocation information amount are 1
Updated every cycle.

Next, as shown in the flow chart of FIG. 21, the band determining means 54 of each hierarchical level obtains the average information of each subband obtained by the average information amount calculating means 51 of each subband every cycle. The amount and the section allocation information amount for one cycle obtained by the information amount control circuit 53 are input, and the section allocation information amount is divided into four to calculate the allocation information amount for each layer, and the allocation is performed for each layer. The band that can be covered by the information amount is calculated from the low band side, and the calculated band is assigned to each layer. Bit allocation circuit 4
At 8, information on the allocated band for one cycle to each layer obtained by the band determining means 54 of each layer and the audible component obtained by the audible component extracting means 42 are input. Re-bit allocation is performed on the audible component in the allocated band of each layer according to the size.

The quantizer 49 quantizes the subband data obtained from the subband n-division filter 41 according to the bit allocation information for each subband obtained from the bit allocation circuit 48. In the hierarchical coding formatter 50, the bit allocation information and the quantized data are formatted for each layer according to the allocated band information of each layer obtained by the band determining means 54 of each layer, and sent to the semiconductor memory 5.

Embodiment 7 FIG. Next, a seventh embodiment of the present invention will be described. FIG. 26 is a block circuit diagram showing the configuration of the hierarchical encoder of the memory audio recorder according to the seventh embodiment. The same reference numerals as in FIG. A low-pass filter for smoothing the information amount for each sub-band calculated by the means 43 to calculate an average information amount for each sub-band; This is a low-pass filter for calculating the amount.

FIG. 27 is a diagram showing the configuration of the information amount control circuit 46 according to the seventh embodiment.
The difference between the information amount per frame according to the desired bit rate obtained from the information amount setting unit 45 and the bit allocation information amount obtained from the bit allocation circuit 48 is extracted. 58 is a low-pass filter for smoothing the extracted difference, 59 is a converter for converting the difference value smoothed by the low-pass filter into an encoding ratio, and 60 is a multiplier, which is calculated by the encoding ratio converter 59. The information amount obtained by the information amount calculation means 44 of each frame is multiplied by the coding ratio K to calculate the allocated information amount.

Next, the operation of the seventh embodiment will be described. In the hierarchical encoder 22, first, the audio signal is divided into n subbands by a subband n-division filter 41, and at the same time, the audio signal is orthogonally transformed by an audible component extracting means 42 into a frequency domain by FFT transform. A masking level based on the auditory characteristics is determined, and an audible component is extracted. As shown in FIG. 16A, the difference between the frequency spectrum and the masking level is the audible component. Further, the audible components are collected for each sub-band by sub-band division, and the audible components for each sub-band are extracted as shown in FIG. Next, the information amount calculating means calculates the information amount for each subband band as shown in FIG. 16C by giving 1-bit information amount to the audible component 6 dB.

Then, the information amount for each sub-band is smoothed by the low-pass filter 55, and the average information amount for each sub-band is obtained. On the other hand, in the instantaneous information amount calculating means 52 for each frame, the information amounts of n subbands are added and 1
The information amount of the frame is calculated. In the information amount per frame setting unit 45 based on the coding rate, the information amount per frame C is calculated based on the bit rate by setting a desired coding rate, and is sent to the information amount control circuit 53. Information amount control circuit 53
Then, according to the average amount of information per frame allocated to the frames encoded so far, the final average allocated amount of information matches the information amount C set by the information amount per frame setter based on the coding rate. In such a manner, the instantaneous allocation information amount is determined for the information amount calculated by the information amount calculation means 44 of each frame.

Then, the instantaneous allocation information amount is smoothed by the low-pass filter 56, and the average allocation information amount is calculated. Next, the band determining means 47 of each layer outputs the low-pass filter 5
5 and the smoothed allocation information amount obtained by the low-pass filter 56 are input, and the average allocation information amount is divided into four parts to thereby divide the average allocation information amount into four. The allocation information amount is calculated, and the band that can be covered by the allocation information amount for each layer is calculated from the low band side, and is set as the allocation band of each layer.

The bit allocation circuit 48 uses the audible component obtained by the audible component extraction means 42 and the audible component obtained by the audible component extraction means 42 to obtain the audible signal within the allocated band of each layer. The components are re-bit-allocated according to their size. Quantizer 49
In accordance with the bit allocation information for each subband obtained from the bit allocation circuit 48, the subband n
The sub-band data obtained from the division filter 41 is quantized. In the hierarchical coding formatter 50,
According to the allocated band information of each layer obtained from the band determining means 47 of each layer, the bit allocation information and the quantized data are formatted for each layer and sent to the semiconductor memory 5.

Next, the information amount control operation in the information amount control circuit 53 will be described. In the information amount control circuit 53, first, the information amount per frame according to the desired bit rate obtained from the information amount setting unit 45 by the comparator 57 and the bit allocation circuit 48
The difference in the obtained bit allocation information amount is extracted, and the difference extracted by the low-pass filter 58 is smoothed. Next, a low-pass filter 58 is output by an encoding ratio converter 59.
Is converted into a coding ratio, and the multiplier 60 multiplies the coding ratio K calculated by the coding ratio converter 59 by the information amount obtained by the information amount calculating means 44 of each frame. Then, the allocation information amount is calculated. With the above operation, the amount of information in variable-length frame coding is controlled on average to a predetermined bit rate by gently controlling the amount of information according to the accumulated value of the difference between the desired bit rate and the allocation information.

As described above, by using hierarchical coding in which each layer code is variable by variable bit allocation and variable band based on the input audio signal, high-quality sound can be efficiently recorded, and the recording time is limited. Thus, a memory audio recorder that can record and reproduce with optimum sound quality can be obtained.

Embodiment 8 FIG. FIG. 28 is a block circuit diagram of Embodiment 8 of the present invention, and the same reference numerals as in FIG. 1 denote the same parts. In the eighth embodiment, a case will be described in which data obtained by converting an audio signal into a frequency domain is quantized. However, data to be quantized may be subband data. In FIG. 28, reference numeral 61 denotes an encoder 6
The digital audio data 63 input to 2 is a time-frequency domain conversion circuit (hereinafter referred to as “frequency domain conversion circuit”), which converts the input digital audio data 61 to the frequency domain. Reference numeral 64 denotes conversion coefficient data converted by the frequency domain conversion circuit 63, and reference numeral 65 denotes a bit allocation circuit, which determines bit allocation to the conversion coefficient data 64 so as to obtain a predetermined sound quality based on characteristics of an input audio signal. 66 is bit allocation information determined by the bit allocation circuit 65, 67 is a quantization circuit, and transform coefficient data 6 obtained by the frequency domain conversion circuit 63.
4 is quantized by the number of bits given by the bit allocation circuit 65. 68 is the quantized data quantized by the quantizing circuit 67, 69 is the bit allocation information 66 and the quantized data 6
A formatting circuit for formatting 8;
70 is a frame length calculation circuit for calculating a frame length based on the bit allocation information 66, 71 is a write address control circuit for controlling a write address according to the frame length calculated by the frame length calculation circuit 70, 63, 65, 6
The encoder 62 is composed of 7, 69, 70 and 71.

Reference numeral 5 denotes a semiconductor memory for storing encoded data; 72, bit allocation information extracted from the semiconductor memory 5; 73, a bit allocation information and frame length calculation circuit which stores the bit allocation information 72 in a temporary buffer. The frame length is calculated from the bit allocation information 72. A read address control circuit 74 controls a read address such that necessary quantized data 75 is read from the semiconductor memory 5 in accordance with the input frame length. Reference numeral 75 denotes quantized data read from the semiconductor memory 5, and 76 denotes an inverse quantization circuit for inversely quantizing the quantized data 75 having the number of bits given by the bit allocation information 72.
Constitute the decoder 77.

FIG. 29 shows bit allocation circuit 6 shown in FIG.
5 shows the configuration of FIG. In FIG. 29, reference numeral 78 denotes a band division energy calculation circuit that divides the coefficient data 64 into a plurality of frequency bands and calculates the energy of each band. 7
Reference numeral 9 denotes an allowable noise level calculation circuit that calculates an allowable noise level of each band based on the energy of each band calculated by the band division energy calculation circuit 78. Reference numeral 80 denotes a bit allocation calculation circuit that determines the number of bits to be allocated to the transform coefficient divided into each band according to the difference between the allowable noise level of each band calculated by the allowable noise level calculation circuit 79 and the energy of each band.

FIG. 30 is a format diagram showing encoded data of one frame which has been highly efficiently encoded by the encoder 62. One frame of coded data includes fixed-length bit allocation information and variable-length quantized data.

FIG. 31 is a diagram showing the relationship between the masking threshold and the minimum audible limit used for bit calculation in the bit allocation circuit 65. FIG. 32 is a diagram illustrating the energy of each band and the calculated allowable noise level.

FIG. 33 shows the case where bit allocation information 66 and quantized data 68 are recorded separately in an auxiliary information recording area and a quantized data recording area when recording encoded data on the semiconductor memory 5. FIG. 9 is a diagram illustrating an operation when performing high-speed playback.

Next, the operation at the time of recording according to the eighth embodiment will be described. The audio signal 61 converted into digital data by the A / D converter 3 is converted into a time signal
The signal is input to the frequency domain conversion circuit 63, and is divided into blocks by a fixed number of samples in the time-frequency domain conversion circuit 63
The conversion coefficient data 64 is converted into the frequency domain on a frame basis, and is input to the bit allocation circuit 65. The bit allocation circuit 65 performs bit allocation based on the characteristics of the input audio signal so as to satisfy a predetermined sound quality. Bit allocation information 6 determined by bit allocation circuit 65
6 is input to the quantization circuit 67. In the quantization circuit 67, the transform coefficient data 64 is quantized based on the input bit allocation information 66 by the number of allocated bits.
The bit allocation information 66 and the quantized data 68 are formatted by a formatting circuit 69. The bit allocation information 66 is input to the frame length calculation circuit 70,
In the frame length calculating circuit 70, the total number of bits allocated to the quantized data is obtained from the number of bits allocated to the transform coefficient and the number of transform coefficients quantized by the number of bits, and transmitted as auxiliary information. The length of one encoded frame is calculated by adding the number of bits of the fixed-length bit allocation information to be written, and sent to the write address control circuit 71.

In this way, by assigning bits so as to obtain a predetermined sound quality and recording the quantized data quantized by the number of bits and the assigned bit allocation information, one frame of a fixed length is recorded. An encoding method that encodes audio data to a variable length is hereinafter referred to as “frame length variable encoding”. The write address control circuit 71 generates a write address according to the input frame length, writes the formatted compressed data into the semiconductor memory 5, and moves the write address according to the frame length. As a result, encoded data is continuously recorded in the semiconductor memory 5 with a variable frame length.

Next, at the time of reproduction, fixed-length bit allocation information 72 is read from the semiconductor memory 5 and stored in the bit allocation information buffer 73. In the bit allocation information buffer 73, the number of bits allocated to the quantized data is calculated from the input bit allocation information, and sent to the read address control circuit 74. Read address control circuit 74
Then, a read address is generated according to the code length of the input quantized data, and the quantized data 75 is read from the semiconductor memory 5. The read quantized data 75 is inversely quantized and decoded in the inverse quantization circuit 76 according to the bit allocation information input from the bit allocation information buffer 73 according to the given number of bits. . The output of the decoder 77 is the D / A converter 8
The audio signal is converted into an analog audio signal, and is output from an audio output terminal 10 through an audio amplifier 9 that amplifies the audio signal to a predetermined audio level.

Next, the bit allocation operation in the bit allocation circuit 65 in the encoder 62 of the semiconductor memory recorder will be described with reference to FIG. The input conversion coefficient data 64 is divided into a plurality of frequency bands by a band division energy calculation circuit 78, and an average energy is calculated from the conversion coefficients of each band. The calculated energy of each band is input to the allowable noise level calculating circuit 79, and the allowable noise level is calculated. In calculating the allowable noise level, a masking effect, a minimum audible limit characteristic, and the like are used in order to perform compression with less deterioration in audibility in consideration of human auditory characteristics. Here, the masking effect is an effect in which a sound in a certain frequency band is masked by a sound having a high level in another frequency band, and the minimum audible limit is a sound of a minimum level that can be heard by a human.

FIG. 31 shows an example in which an allowable noise level is calculated based on the masking effect and the minimum audibility for the energy of each band. In FIG. 31, S1 to S10 indicate the energy of each frequency band when divided into ten frequency bands. The diagonal lines extending left and right from the energy of each band
The area masked by the sound in that band is shown. The characteristic curve indicated by the broken line indicates the minimum audible limit. As the allowable noise level of each frequency band, the maximum noise level is selected from the masking levels from other frequency bands and the minimum audible level. In the example of FIG. 31, the allowable noise level of each selected frequency band is indicated by a horizontal line in FIG. The allowable noise level calculated by the allowable noise level calculation circuit 79 and the energy of each band are input to the allocation bit calculation circuit 80.
Since the allocated bit calculation circuit 80 does not hear sounds below the allowable noise level, the number of bits according to the difference between the energy of each frequency band and the allowable noise level is calculated for the sound in the band exceeding the allowable noise level, and the number of bits is calculated. The assignment information 66 is output. The number of bits allocated in this way is based on the properties of the input audio signal, and the number of bits allocated for encoding one frame is variable.

As described above, the coded data of one frame to which bits have been allocated is composed of fixed-length bit allocation information and variable-length quantized data quantized by the allocated number of bits, as shown in FIG. And formatted. By using the fixed-length bit allocation information as the auxiliary information as described above, the number of bits allocated to the quantized data can be calculated, so that encoding in a variable-length frame is possible. Further, in the semiconductor memory 5, the recording position can be randomly determined at high speed by controlling the address, so that the code of the variable length frame can be recorded.

Also, the code of the variable length frame is not recorded in the format as shown in FIG. 30, but an area for recording auxiliary information and an area for recording quantized data are provided on the semiconductor memory 5 to record the auxiliary information. Bit allocation information 66 is continuously recorded at a fixed length in an area where
The variable length quantized data is continuously recorded in the area where 8 is recorded. Thus, the address at which the quantized data is recorded is calculated based on the fixed-length bit allocation information in the auxiliary information area, and the quantized data is read out and decoded at a predetermined frame interval, thereby enabling high-speed reproduction.

FIG. 33 shows an example of this. In the example of FIG. 33, 999 is an auxiliary information (bit allocation information) recording area from address 0, and a quantized data recording area is from address 1000. Bit allocation information is recorded in 20 bits from address 0. The number indicated in the bit allocation information is the number of bits of one frame of the quantized data calculated from the bit allocation information, and the quantized data indicated by the number is variable in the quantized data recording area. It is recorded continuously in long. The numbers circled in the figure indicate frame numbers. Therefore, the number in the bit allocation information is replaced with the first address 1000 of the quantized data recording area.
Is added, the first address where the quantized data is recorded can be calculated. Therefore, first, the bit allocation information in the auxiliary information is read, the address where the quantized data is recorded is calculated from the read bit allocation information, and the quantized data is read from the address calculated every five frames. By decoding, high-speed reproduction of 5 times speed becomes possible. The reading of the quantized data is performed using frames 1, 6, 11, 16,...
The addresses of frames to be read out and read in the order of 1000, 1378, 1723, 2112,.
Becomes

As described above, a variable frame length encoding method for allocating bits with variable frame lengths to obtain a predetermined sound quality based on an input audio signal, and encoding the bits into bit allocation information and quantized data. By using and storing the variable-length frames in a semiconductor memory capable of high-speed random access, a semiconductor memory recorder capable of efficiently recording with high sound quality and capable of recording for a long time is obtained.

Embodiment 9 FIG. Next, a ninth embodiment of the present invention will be described with reference to the drawings. In FIG. 34, the same reference numerals as those in FIG. 28 denote the same parts, respectively. Reference numeral 81 denotes a reproduction frame selection circuit which calculates the number of bits allocated to the quantized data based on the bit allocation information. Only a frame exceeding a predetermined threshold is selected as a reproduction frame. Reference numeral 82 denotes reproduction frame selection information and reproduction frame bit allocation information obtained by the reproduction frame selection circuit 81, and reference numeral 83 denotes a reproduction frame selection circuit 8.
It is the code length of the quantized data calculated in step 1 and the reproduction frame selection information.

The operation at the time of recording according to the ninth embodiment is the same as that of the first embodiment, and therefore the description is omitted. Next, the “quick listening operation” during reproduction will be described. "Early listening" here
This means that when a semiconductor memory recorder is used for recording conversations such as meeting memos, it is not a high-speed playback by continuously skipping data at a fixed interval or a fast listening by increasing the playback speed. By skipping the part where the conversation is not recorded only by the noise of the utterance, it is possible to perform the early listening by continuously reproducing only the necessary conversation part.

The fixed-length bit allocation information 72 is read from the semiconductor memory 5 and input to the reproduction frame selection circuit 81. The reproduction frame selection circuit 81 calculates the number of bits allocated to one frame of quantized data based on the bit allocation information 72. Here, when bits are allocated by the method as described in the eighth embodiment, the number of bits allocated to the quantized data is 0 since all transform coefficients are at a level below the minimum audible level in a silent part. Becomes

Further, even in a portion where only noise is recorded, the number of bits allocated to the quantized data is very small because the overall level is small. Therefore, a predetermined threshold value is provided for the number of bits allocated to one frame, and a frame in which the allocated number of bits does not exceed the threshold value is not reproduced as a silent portion or a noise portion. The reproduction frame selection information and the bit allocation information 82 are transmitted to the inverse quantization circuit 76 for the frames exceeding the limit.

The code length and the reproduction frame selection information 83 for one frame of the quantized data calculated by the reproduction frame selection circuit 81 are read from the read address control circuit 7.
4 When a frame selected as a reproduction frame is input, the read address control circuit 74 generates a read address in accordance with the code length of the quantized data that has been input, and the quantized data 75 is output from the semiconductor memory 5. Is read.

In a frame not selected as a reproduction frame, the address is shifted by the number of bits assigned to the quantized data. Read quantized data 7
In 5, the quantized data of the given number of bits is inversely quantized and decoded in the inverse quantization circuit 76 in accordance with the bit allocation information 82 input together with the frame selection information. The output of the decoder 77 is converted into an analog audio signal by the D / A converter 8 and output from the audio output terminal 10 through the audio amplifier 9 for amplifying the audio signal to a predetermined audio level.

As described above, in the variable frame length coding in which bit allocation is performed based on an input signal, frames with a small amount of information, such as silent portions, have very few bit allocations. By reproducing only the frames exceeding the threshold value, the silent section and the like can be skipped, and only the portion where the voice is recorded can be played back at a high speed.

[0124]

As described above, according to the present invention, it is possible to use the hierarchical coding method which gives priority to the encoding of the audio signal and make full use of the random access property of the semiconductor memory of the recording medium. Thus, the recording time of the semiconductor memory audio recorder is switched, so that the recording is always performed at the optimum sound quality regardless of the recording time specified for the recording medium (when the recording time is short, the recording is performed at the highest sound quality, (Sound quality deteriorates as the length increases.) There is an effect that a semiconductor memory audio recorder capable of recording and reproducing can be obtained.

Also, the use of hierarchical coding with priority in audio signal coding and the random accessibility of a semiconductor memory as a recording medium make it possible to reduce the recording time of a semiconductor memory recording / reproducing apparatus. Because it is configured to be flexible, it always has the optimal sound quality regardless of the recording time specified for the recording medium (high-quality sound is recorded when the recording time is short, and sound quality deteriorates as the recording time becomes longer). There is an effect that a semiconductor memory audio recorder capable of recording and reproduction and variable speed reproduction can be obtained.

Also, the use of hierarchical coding with priority for audio signal coding and the method of hierarchical coding are described by using a variable-length frame code for performing optimal bit allocation according to the information amount of each layer. In order to make the recording time of the semiconductor memory recording / reproducing device flexible by using the same amount of information in each layer while using the optimization, and making full use of the random accessibility of the semiconductor memory as a recording medium. With this configuration, recording and reproduction can always be performed at an optimum sound quality (when the recording time is short, recording is performed at a high sound quality, and when the recording time is prolonged, the sound quality deteriorates) regardless of the recording time determined for the recording medium, and There is an effect that a semiconductor memory audio recorder capable of variable speed reproduction can be obtained.

Further, based on the characteristics of the input audio signal, bit allocation is performed so as to maintain a predetermined sound quality, bit allocation information is recorded as auxiliary information, and a frame length variable which can be encoded in a variable length frame from this bit allocation information. By using the encoding method and controlling the address for accessing the semiconductor memory using the bit allocation information,
By making full use of the high-speed random access property of the semiconductor memory of the recording medium and efficiently recording on the semiconductor memory audio recording / reproducing apparatus, there is an effect that a semiconductor memory audio recording / reproducing apparatus having high sound quality and a long recording time can be obtained.

Further, the auxiliary information recording area and the quantized data recording area are provided on the semiconductor memory, and the bit allocation information is continuously recorded at a fixed length in the auxiliary information recording area. There is an effect that becomes possible.

Also, since only frames whose information amount exceeds a predetermined threshold value are reproduced by using the bit allocation information by the variable frame length encoding, "early listening" can be realized with a very simple circuit configuration. There is an effect that a possible semiconductor memory audio recording / reproducing device can be obtained.

[Brief description of the drawings]

FIG. 1 is a block circuit diagram of a semiconductor memory audio recorder according to Embodiment 1 of the present invention.

FIG. 2 is a diagram illustrating a configuration example of a hierarchical coding scheme according to the first embodiment.

FIG. 3 is a diagram illustrating another configuration example of the hierarchical coding scheme according to the first embodiment.

FIG. 4 is a memory map of the semiconductor memory according to the first embodiment;

FIG. 5 is a diagram showing a recording process of an audio signal in the semiconductor memory according to the first embodiment.

FIG. 6 is a block circuit diagram of a semiconductor memory audio recorder according to Embodiment 2 of the present invention.

FIG. 7 is a block circuit diagram of a semiconductor memory audio recording / reproducing apparatus according to Embodiment 3 of the present invention.

FIG. 8 is a diagram for explaining the concept of hierarchical coding according to the third embodiment.

FIG. 9 is a frequency characteristic diagram showing hierarchical levels of hierarchical encoding according to the third embodiment.

FIG. 10 is a diagram illustrating a configuration of a hierarchical encoder according to a third embodiment and a concept of a recording method in a semiconductor memory.

FIG. 11 is a memory map of the semiconductor memory according to the third embodiment;

FIG. 12 is a block circuit diagram of a semiconductor memory audio recording / reproducing apparatus according to Embodiment 4 of the present invention.

FIG. 13 is a diagram illustrating a relationship between an audio decoding time and an audio reproduction time according to the fourth embodiment.

FIG. 14 is a frequency characteristic diagram showing hierarchical levels of hierarchical encoding of the semiconductor memory audio recording and reproducing devices according to the fifth to seventh embodiments of the present invention.

FIG. 15 is a diagram illustrating a configuration of a hierarchical encoder according to a fifth embodiment and a concept of a recording method in a semiconductor memory.

FIG. 16 is a diagram illustrating a state of extraction of an audible component and calculation of an information amount in the hierarchical encoder according to the fifth to seventh embodiments.

FIG. 17 is a diagram illustrating a manner of determining an allocated band of each layer in the layer encoders according to the fifth to seventh embodiments.

FIG. 18 is a diagram illustrating a configuration of a hierarchical encoder according to a sixth embodiment.

FIG. 19 is a flowchart illustrating an average information amount calculation operation for each subband by an information amount calculation unit according to the sixth embodiment.

FIG. 20 is a flowchart showing an operation of calculating a section allocation information amount in the information amount control means of the sixth embodiment.

FIG. 21 is a flowchart illustrating an allocated band determining operation in a band determining unit of each layer according to the sixth embodiment.

FIG. 22 is a flowchart illustrating a bit allocation operation in the bit allocation circuit according to the sixth embodiment.

FIG. 23 is a diagram showing the relationship between the maximum band SBmax in which an audible component exists and the information amount in the instantaneous information amount determining means of the sixth embodiment.

FIG. 24 is a diagram showing the relationship between the maximum band SBmax in which an audible component exists and the information amount in the instantaneous information amount determining means according to the sixth embodiment.

FIG. 25 is a diagram showing the relationship between the maximum band SBmax in which an audible component exists and the information amount in the instantaneous information amount determining means of the sixth embodiment.

FIG. 26 is a diagram showing a configuration of a hierarchical encoder of a semiconductor memory audio recording / playback device according to a seventh embodiment of the present invention.

FIG. 27 is a diagram illustrating a configuration of an information amount control circuit of the hierarchical encoder according to the seventh embodiment.

FIG. 28 is a block circuit diagram of a semiconductor memory audio recording / reproducing device according to an eighth embodiment of the present invention.

FIG. 29 is a diagram illustrating a configuration of a bit allocation circuit of the encoder according to the eighth embodiment.

FIG. 30 is a diagram illustrating a format of an encoded frame according to the variable frame length encoding method according to the eighth embodiment.

FIG. 31 is a diagram illustrating a state of calculation of an allowable noise level using a masking effect and a minimum audibility in the bit allocation circuit of the encoder according to the eighth embodiment.

FIG. 32 is a diagram illustrating a state of an allowable noise level and energy of each band in a bit allocation circuit of an encoder according to an eighth embodiment.

FIG. 33 is a memory map for describing high-speed playback according to the eighth embodiment.

FIG. 34 is a block circuit diagram of a semiconductor memory audio recording / reproducing apparatus according to Embodiment 9 of the present invention.

[Explanation of symbols]

3 A / D converter, 4 layer encoder, 5 semiconductor memory, 6 memory address controller, 7 layer decoder, 8
D / A converter, 11 hierarchical level identification code generator,
12 hierarchy level identification code regenerator, 13 memory capacity detector, 15 subband split filter, 16 bit allocator, 17 first semiconductor memory, 18 second semiconductor memory, 19 hierarchy level converter, 20 first memory Address controller, 21 second memory address controller 22 hierarchical encoder, 23 memory address controller, 2
4 layer decoder, 25 layer level identification code generator, 26 layer level identification code reproducer, 27 division filter, 28 MDCT, 29 block size setting unit, 30 grouper, 31 layerer / quantizer,
32 dynamic bit allocator, 33 scale factor calculator, 34 formatter, 35 address switcher, 36 write address generator, 37 read address generator, 38 hierarchical level determiner, 39
Clock divider, 40 Playback speed setting device, 41
Subband n division filter, 42 audible component extraction means,
43 information amount calculating means for each subband of each frame,
44 information amount calculation means for each frame; 45 information amount setting device per frame based on coding rate; 46 information amount control circuit; 47 band determination means for each layer;
1 average information amount calculation means for each subband, 52 instantaneous information amount calculation means for each frame, 53 information amount control circuit, 54 band determination means for each layer, 55 information amount smoothing filter for each subband, 56 allocation information Amount smoothing filter, 57 information amount comparator, 58 difference information amount smoothing filter, 59 coding ratio converter, 62 encoder, 65 bit allocation circuit, 67 quantization circuit, 70
Frame length calculation circuit, 71 write address control circuit, 73 bit allocation information buffer and frame length calculation circuit, 74 read address control circuit, 77 decoder, 78 band division energy calculation circuit, 79 allowable noise level calculation circuit, 80 allocation Bit calculation circuit, 81
Reproduction frame selection circuit, 82 Reproduction frame selection information and bit allocation information, 83 Reproduction frame selection information and code length of quantized data.

 ──────────────────────────────────────────────────続 き Continued on the front page (31) Priority claim number Japanese Patent Application No. 5-1850 (32) Priority date January 7, 1993 (1.7.1.7) (33) Priority claim country Japan (JP) (72) Inventor Yoshinobu Ishida 2-3-2 Marunouchi, Chiyoda-ku, Tokyo F-term in Mitsubishi Electric Corporation (reference) 5D045 DA20 5J064 AA01 BC01 BC02 BC06 BC07 BC16 BC21 BD01

Claims (15)

[Claims] 1. A frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and dividing the obtained conversion coefficient in accordance with a level change based on psychoacoustic characteristics to provide a human auditory sense. N (n is a natural number of 2 or more) first priority-ordered based on characteristics
An audio signal encoding apparatus, comprising: a layering / quantizing means for generating the n-th encoded data. 2. A frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and dividing the obtained conversion coefficient by using a frequency level and a quantization level as parameters to obtain a human auditory sense. N (n is a natural number of 2 or more) with priorities based on characteristics
An audio signal encoding apparatus comprising a layering / quantizing means for generating the first to n-th encoded data. 3. A frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and performing hierarchization and quantization according to the following rules on the obtained conversion coefficient. n (n is a natural number of 2 or more)
An audio signal encoding apparatus, comprising: a layering / quantizing means for obtaining encoded data divided into hierarchical levels. 1. Among the transform coefficients, a transform coefficient whose frequency band is a transform coefficient up to a predetermined frequency f1 and whose quantization level is from the MSB side to a predetermined number of bits b1 is selected, and is selected as a hierarchical level 1 Coded data S1. 2. Among the conversion coefficients, the frequency band is a conversion coefficient up to a predetermined frequency f2 (f2 ≧ f1), and
The quantization level is a predetermined number of bits b2 from the MSB side.
The transform coefficients up to (b2 ≧ b1) are selected, and the residual signal obtained by subtracting the transform coefficient of the hierarchical level 1 from this signal is used as the encoded data S2 of the hierarchical level 2. 3. Of the transform coefficients, the frequency band is a transform coefficient up to a predetermined frequency fn (fn ≧ fn−1), and the quantization level is a predetermined number of bits b from the MSB side.
n (bn.gtoreq.bn-1), and further, from this signal, the hierarchical levels 1 to n-
The residual signal obtained by subtracting the transform coefficient of 1 is used as the coded data Sn of the hierarchical level n. 4. A frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and extracting an audible signal component based on psychoacoustic characteristics from the obtained conversion coefficient. Audio signal encoding characterized by comprising layering / quantizing means for performing layering and quantization in accordance with a rule to obtain encoded data divided into n (n is a natural number of 2 or more) hierarchical levels. apparatus. 1. Among the transform coefficients of the audible signal component, a transform coefficient whose frequency band is up to a predetermined frequency f1 is selected, and this is defined as hierarchical level 1 encoded data S1. 2. From among the transform coefficients of the audible signal component, a transform coefficient whose frequency band is up to a predetermined frequency f2 (f2 ≧ f1) is selected. It is assumed that the encoded data S2 is at the hierarchical level 2. 3. From among the transform coefficients of the audible signal component, transform coefficients whose frequency band is up to a predetermined frequency fn (fn ≧ fn−1) are selected, and further, from this signal, the transform of the hierarchical levels 1 to n−1 is performed. The residual signal from which the coefficient has been subtracted is coded data Sn of the hierarchical level n. 5. The audio signal encoding apparatus according to claim 2, wherein the division into the hierarchical levels is performed in association with a level change based on psychoacoustic characteristics. 6. The apparatus according to claim 1, further comprising a dividing unit for dividing the input digital audio signal into blocks of a predetermined size, wherein the frequency converting unit performs a conversion on each of the divided blocks. 5. The audio signal encoding device according to claim 4. 7. Hierarchical coded audio data to which hierarchical priorities based on human auditory characteristics are given by dividing the frequency-converted transform coefficients according to level changes based on auditory psychological characteristics. By inputting the identification code at the hierarchical level of the hierarchically encoded audio data as input and decoding the hierarchically encoded audio data according to the hierarchical level based on the identification code, the frequency An audio signal decoding apparatus comprising: decoding means for obtaining a transform coefficient corresponding to a band; and frequency inverse transform means for performing an inverse transform on the obtained transform coefficient to obtain an original digital audio signal. . 8. A hierarchically encoded audio data to which hierarchical priorities based on human auditory characteristics are given by dividing the frequency-transformed transform coefficients using the level of frequency and the level of quantization level as parameters. Inputting the hierarchically encoded audio data and the identification code at the hierarchical level of the hierarchically encoded audio data, and decoding the hierarchically encoded audio data according to the hierarchical level based on the identification code. Audio signal decoding comprising: decoding means for obtaining a transform coefficient corresponding to a frequency band; and frequency inverse transform means for performing an inverse transform on the obtained transform coefficient to obtain an original digital audio signal. Device. 9. Hierarchical coded audio data that has been hierarchized and quantized according to the following rules and has been given a hierarchical priority based on human auditory characteristics, and said hierarchical coding By inputting the identification code at the hierarchical level of the audio data as input, the hierarchically encoded audio data is decoded according to the hierarchical level based on the identification code, so as to correspond to the frequency band. An audio signal decoding apparatus comprising: decoding means for obtaining a transform coefficient; and frequency inverse transform means for performing an inverse transform on the obtained transform coefficient to obtain an original digital audio signal. 1. Among the conversion coefficients obtained by frequency-converting the digital audio signal, the frequency band is a conversion coefficient up to a predetermined frequency f1, and the quantization level is a conversion coefficient from the MSB side to a predetermined bit number b1. A coefficient is selected, and this is set as encoded data S1 of the hierarchical level 1. 2. Among the conversion coefficients, the frequency band is a conversion coefficient up to a predetermined frequency f2 (f2 ≧ f1), and
The quantization level is a predetermined number of bits b2 from the MSB side.
The transform coefficients up to (b2 ≧ b1) are selected, and the residual signal obtained by subtracting the transform coefficient of the hierarchical level 1 from this signal is used as the encoded data S2 of the hierarchical level 2. 3. Of the transform coefficients, the frequency band is a transform coefficient up to a predetermined frequency fn (fn ≧ fn−1), and the quantization level is a predetermined number of bits b from the MSB side.
n (bn.gtoreq.bn-1), and further, from this signal, the hierarchical levels 1 to n-
The residual signal obtained by subtracting the transform coefficient of 1 is used as the coded data Sn of the hierarchical level n. 10. Hierarchical coded audio data that has been hierarchized and quantized in accordance with the following rules, and given a hierarchical priority based on human auditory characteristics, and said hierarchical coding By inputting the identification code at the hierarchical level of the audio data as input and decoding the hierarchically encoded audio data according to the hierarchical level based on the identification code, it is possible to cope with the frequency band. An audio signal decoding device comprising: decoding means for obtaining a transform coefficient; and frequency inverse transform means for performing an inverse transform on the obtained transform coefficient to obtain an original digital audio signal. 1. Of the transform coefficients of the audible signal component extracted based on the human psychoacoustic characteristics, the frequency band of the transform coefficient is a predetermined frequency f
The transform coefficients up to 1 are selected, and this is used as the encoded data S1 of the hierarchical level 1. 2. From among the transform coefficients of the audible signal component, a transform coefficient whose frequency band is up to a predetermined frequency f2 (f2 ≧ f1) is selected. Hierarchical level 2 encoded data
Data S2. 3. From among the transform coefficients of the audible signal component, transform coefficients whose frequency band is up to a predetermined frequency fn (fn ≧ fn−1) are selected, and further, from this signal, the transform of the hierarchical levels 1 to n−1 is performed. The residual signal from which the coefficient has been subtracted is coded data Sn of the hierarchical level n. 11. A frequency conversion means for converting an input digital audio signal into a conversion coefficient corresponding to a frequency band, and said converting means for obtaining a predetermined sound quality at the time of reproduction according to characteristics of said digital audio signal. Bit allocation means for obtaining allocation information of bits allocated to coefficients; quantization means for obtaining variable-length quantized data by quantizing the transform coefficient based on the bit allocation information; And the fixed-length bit allocation information,
Formatting means for obtaining variable-length frame data; frame-length detecting means for obtaining a frame length of the frame data; and an address for writing to a recording medium according to the frame length. An audio signal encoding device comprising: an address control unit for controlling the audio signal. 12. A bit allocation means, comprising: a band division energy detecting means for obtaining energy of a transform coefficient of each band obtained by dividing a transform coefficient according to a frequency band; and an allowance of each band based on the energy of each band. An allowable noise level detecting means for obtaining a noise level; and dividing into respective bands based on a difference between an energy of each band obtained by the band dividing energy detecting means and an allowable noise level of each band obtained by the allowable noise level detecting means. 12. The audio signal encoding apparatus according to claim 11, further comprising an allocation bit detection unit for obtaining the number of bits allocated to the transformed coefficient. 13. A sub-band dividing means for dividing an input digital audio signal into a plurality of sub-band data, and each sub-band data according to the characteristics of the digital audio signal so as to obtain a predetermined sound quality during reproduction. Bit allocation means for obtaining bit allocation information to be allocated to the data; quantizing means for obtaining variable-length quantized data by quantizing each of the sub-band data based on the bit allocation information; Data and the fixed-length bit allocation information,
Formatting means for obtaining variable-length frame data; frame-length detecting means for obtaining a frame length of the frame data; and a recording medium for recording on a recording medium according to the frame length. An audio signal encoding device comprising: an address control unit that controls a write address. 14. An audio signal decoding system comprising quantized data quantized to a variable length according to characteristics of an audio signal and fixed-length bit allocation information for determining quantization of the quantized data. An apparatus for extracting bit allocation information of a fixed length from the frame data, and obtaining bit allocation information and a frame length of the frame data from the obtained bit allocation information. Detecting means for extracting quantized data quantized to a variable length from the frame data based on the obtained frame length, and extracting the extracted quantized data into an inverse quantum based on the bit allocation information; An audio signal decoding device comprising: 15. A frame data comprising quantized data quantized to a variable length according to characteristics of an audio signal and fixed-length bit allocation information for determining quantization of the quantized data. An audio signal decoding device for extracting and decoding quantized data quantized to a variable length from data, wherein fixed-length bit allocation information is extracted from the frame data. The number of bits allocated to quantized data quantized to a variable length is obtained from the allocated bit information, and the reproduction frame selection for selecting frame data in which the number of bits exceeds a predetermined threshold value. Means for extracting quantized data quantized to variable length from the selected frame data, and dequantizing means for dequantizing the extracted quantized data based on the bit allocation information. An audio signal decoding device characterized in that: