JPH075896A - Data compressing and expanding method - Google Patents

Abstract

PURPOSE:To reproduce high quality digital voice signals by restoring real and imaginary parts of frequency components based on data, performing an inverse orthogonal transformation to the frequency components and restoring to digital audio signals. CONSTITUTION:A data compression device 10 is used for audio data transmission or recording ans a data expansion device 20 is used for music data receiving or reproducing. The transmitting or the recording sides respectively divide the real and the imaginary parts of each frequency components, that are obtained through the orthogonal transformation, plural bands and the power of 2, which is larger than the maximum value of the absolute value of respective frequency components and is closest to the maximum value, is decided as the scale for normalization for every divided band. Based on this scale, the real and the imaginary parts of the frequency components are normalized and the normalizing scale data, the absolute values of the normalized data and the code bits are transmitted or recorded. Moreover, the receiving or the reproducing sides reproduce the real and the imaginary parts of the frequency components based on the data and an inverse orthogonal transformation is performed to the frequency components obtained and digital audio signals are reproduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data compression / decompression method for compressing a digital audio signal on the transmitting or recording side and expanding it on the receiving or reproducing side to restore the audio signal.

[0002]

2. Description of the Related Art As a coding technique for efficiently utilizing a limited transmission band to transmit a digital signal, for example, a book "Digital Audio Dictionary", pages 18 to 19 (published by Ohmsha, Inc.) High-efficiency coding "No. 12-1
P. 5 (Kazuo Nakata, published by Morikita Publishing), "Digital Speech Processing, Coding in Frequency Domain", pages 109 to 113 (published by Tokai University Press), and put into practical use.

In particular, recently, band division coding (Sub Ban
A method of realizing highly efficient compression by taking human auditory characteristics into d Coding (SBC), and transmitting and recording has been widely used. These technologies are, for example,
Books "High-efficiency coding of speech", pages 56-66 (Kazuo Nakata, published by Morikita Publishing), "High-efficiency coding technology for DCC", pages 6-9 (JAS Conference '92 Proceedings),
"High-efficiency coding technology", pages 112 to 115 (JAS Conference '92 Proceedings), JP-A-2-501507, "Digital Adaptive Transform Coding Method".

[0004]

Although the above-mentioned conventional technique certainly enables high-efficiency compression, for example, 20
The signal digitized by the bit AD converter,
DAT and CD as conventional 16-bit recording media
However, there is a problem in that the original 20-bit signal cannot be restored at the time of reproduction by compressing and storing it in a memory, etc., which may impair the musicality of the music signal. This is a similar problem in the signal transmission system.

The present invention has been made to solve the above-mentioned problems, and it is possible to efficiently use a transmission system or a recording medium for a digital audio signal, and a high-quality digital audio signal. It is an object of the present invention to provide a data compression / decompression method that enables decompression.

[0006]

According to a data compression / decompression method of the present invention, a transmitting or recording side orthogonally transforms a digital audio signal, and a plurality of real and imaginary parts of each obtained frequency component are respectively provided. , The power of 2 that is greater than the maximum absolute value of the frequency component and closest to the maximum value is determined for each of the divided bands as a normalization scale, and the real part of the frequency component is determined according to this scale. And normalize the imaginary part, and transmit or record the normalized scale data, the absolute value of the normalized data and its sign bit, and the receiving or reproducing side, the normalized scale data, the absolute value of the normalized data And the real number part and the imaginary number part of the frequency component based on the sign bit, and inverse orthogonal transform is performed on the obtained frequency component to restore the digital audio signal.

In the data compression / decompression method according to the second aspect, when the data amount corresponding to a predetermined number of samples exceeds a predetermined value, the highest frequency component is transmitted or recorded as zero sequentially until the predetermined amount is reached. .

According to a third aspect of the data compression and expansion method of the present invention, the transmitting or recording side orthogonally transforms the digital audio signal and divides the obtained real number part and imaginary number part of each frequency component into a plurality of bands. Normalize with a predetermined normalization scale for each divided band, and send the normalization scale data or a flag indicating that it is a predetermined normalization scale, the absolute value of the normalized data and its sign bit, or A flag indicating that the recorded or received or reproduced side is the normalization scale data or a predetermined normalization scale, the absolute value of the normalized data and the real part and imaginary number of the frequency component based on the sign bit thereof. It is characterized in that a digital voice signal is obtained by restoring the part and subjecting the obtained frequency component to inverse orthogonal transform.

According to a fourth aspect of the data compression and expansion method of the present invention, the transmitting or recording side removes a frequency component of one frequency region which is symmetrical with respect to a predetermined frequency as a boundary, and transmits or records, receives or reproduces. The side also restores the frequency components of the removed frequency domain by utilizing the symmetry of the frequency components.

[0010]

The principle of the present invention will be described below with reference to FIGS. FIG. 3A shows continuous 1024 samples of the music signal whose signal level is close to the maximum, in which the horizontal axis represents time and the vertical axis represents the audio signal level. Here, the voice signal has a maximum positive value "+" centered on the reference level "0".
It continuously changes between "1.0" and the maximum negative value "-1.0". When this music signal is subjected to, for example, a Discrete Fourier Transform, a real part of the frequency component shown in FIG. 3B and an imaginary part of the frequency component shown in FIG. 3C are obtained. In FIGS. 3B and 3C, the horizontal axis represents frequency and the vertical axis represents logarithmic value with the voice level "1.0" set to "0" decibel. Further, FIG. 3 (d) is a logarithmic value of the absolute value of the voice level in which the voice “1.0” of FIG. 1 is set to “0” decibel, and the horizontal axis is the same time as in FIG. 3 (a).

In FIG. 3, the real part and the imaginary part of the frequency component are all "-20" decibels or less, but the decibel value of the input signal is "-9" from a value close to "0" decibel.
It varies in the range of "0" decibels. If one component of the digital audio signal shown in FIG. 3 (d) is encoded with 20 bits, all of these 20 bits must be transmitted or recorded, whereas in FIG. ) And the frequency components shown in FIG. 3 (c) are respectively encoded, at least the upper 3 bits of 20 bits are made unnecessary and can be transmitted or recorded by encoding with 17 bits. Also,
In the real and imaginary parts of the frequency components shown in FIGS. 3 (b) and 3 (c), the maximum level gradually decreases as the frequency region increases. This means that high frequency components in the frequency domain can be encoded with a smaller number of bits.

Therefore, in the data compression / decompression method according to the first aspect, the real number part and the imaginary number part of each frequency component obtained by orthogonal transform are divided into a plurality of bands, and the frequency component is divided into each divided band. The power of 2 that is larger than the maximum absolute value of and that is closest to the maximum is determined as the normalization scale, and the real part and the imaginary part of the frequency component are normalized according to this scale. The absolute value of the converted data and its sign bit are transmitted or recorded. As a result, it is possible to reduce the number of bits per sample as compared with the case where the time-domain audio signal is directly encoded and transmitted or recorded, and the number of bits can be further reduced in a higher frequency region. Therefore, the transmission system or the recording medium can be efficiently used.

Further, in the data compression / decompression method according to the first aspect, when the normalized scale data, the absolute value of the normalized data and the sign bit thereof are transmitted or recorded, the receiving or reproducing side converts the data into these data. Based on this, the real and imaginary parts of the frequency component are restored, and the obtained frequency component is subjected to inverse orthogonal transformation to restore the digital audio signal, so that it is possible to restore to a high-quality digital audio signal.

FIGS. 3 (a) to 3 (d) show 1 in the time domain of the music signal.
Although the logarithmic value of the frequency component and the instantaneous level in the frequency domain for 024 samples is shown, the frequency component and the instantaneous level of the frequency domain for the 16 samples in the time domain of the music signal are also shown for easy understanding. The logarithmic values are shown in Fig. 4 (a)-(d). That is,
FIG. 4 (a) shows 16 consecutive samples of a music signal whose voice level is close to the maximum. The horizontal axis represents time and the vertical axis represents the voice signal level, centered around the reference level "0". Positive maximum value "+1.0" and negative maximum value "-1.0"
Is continuously changing between and. When this music signal is subjected to the discrete Fourier transform, the real part of the frequency component shown in FIG. 4B and the imaginary part of the frequency component shown in FIG. 4C are obtained. In FIGS. 4B and 4C, the horizontal axis represents frequency and the vertical axis represents logarithmic value with the voice level "1.0" set to "0" decibel. 4 (d) is a logarithmic value with the absolute value "1.0" of the sound level of FIG. 4 (a) being "0" decibels, and the horizontal axis is the same time as in FIG. 4 (a). .

4 (a) to 4 (d), the time domain music signal, the real and imaginary parts of the frequency component obtained by orthogonally transforming the signal, and the inverse orthogonal transform are obtained. The values are shown in Table 1 below.

[0016]

[Table 1]

Here, the imaginary part of the frequency component is "+0.0."
When the upper and lower values are compared based on the position "0000", the real part of the frequency component is the nth (n = 1 to 7) counted upward from the reference position, It can be seen that the sign is opposite to the n-th one counted downward and the numerical values are equal to each other. Further, in the imaginary part of the frequency component, it can be seen that both the sign and the numerical value are the same for the nth counted upward from the reference position and the nth counted downward. That is, when these values are displayed in two-dimensional coordinates, the value of the real part is point contrast, and the value of the imaginary part is line symmetrical. This means that when transmitting or recording the real part and the imaginary part, either the upper frequency region or the lower frequency region with the reference point as the boundary, the transmitting or recording side does not need to transmit or record the data. By utilizing the symmetry, the data in the other frequency domain can be predicted.

By the way, a normal music signal has a frequency characteristic as described above, but since a synthesizer sound has a frequency characteristic different from this, even if a normalization scale is determined for each divided band, the scale Large data may continue. In this case, the amount of data to be transmitted or recorded, that is, the amount of compressed data will be much larger than expected.

A data compression / decompression method according to a second aspect deals with such a case, and when the data amount corresponding to a predetermined number of samples exceeds a predetermined value, there is a need until the data amount reaches the predetermined value. A relatively low component, that is, a component having the highest frequency is transmitted or recorded as zero sequentially.

On the other hand, when it is known that the digital music signal does not include a special musical instrument sound, it is not necessary to determine the normalization scale for each divided band and record or send the normalization scale. If the coding is performed using the coding scale described above, the processing time of transmission / reception or recording / reproduction can be shortened, and it becomes unnecessary to send the scale information for each block unit of orthogonal transformation (for example, for every 1024 samples). The data compression / decompression method according to claim 3 focuses on this characteristic, and normalizes by a predetermined normalization scale for each divided band, normalization scale data, normalization absolute value data, and its code. Data is being transmitted or recorded.

In the data compression / decompression method according to the fourth aspect, in consideration of the symmetry of the frequency components, one of the symmetrical components is removed and absolute value data is transmitted or recorded, whereby the transmission / reception data or the recording is performed. The reproduction data can be reduced to about half, and the transmission system or the recording medium can be used more efficiently.

[0022]

The present invention will be described in detail below with reference to the embodiments shown in the drawings. FIG. 1 is a block diagram showing the configuration of a data compression / decompression device embodying the present invention. Here, the data compression device 10 is used for transmitting or recording music data, and the data decompressing device 20 is used for receiving or reproducing music data.

Of these, the data compression apparatus 10 includes an input buffer 11 which holds an input digital signal for a short time,
A memory 12 for storing a digital signal input via the buffer 11, a digital signal processing unit (hereinafter referred to as a DSP) 13 for converting time axis data stored in this memory into frequency axis data, and The memory 14 for storing frequency components and the frequency axis data stored in this memory are used, for example, to determine a normalization scale for each division band equally divided on the frequency axis, and a value of "0 DSP15 which removes the upper bit which becomes, and produces the absolute value of the data, its sign bit and the scale data.
And an encoding circuit 16 for transmitting these data are connected by a common bus 17.

The data decompression device 20 also includes a decoding circuit 2 for decoding the data sent from the data compression device.
1, a memory 22 for storing the decoded data, a DSP 23 for restoring the real and imaginary parts of the frequency component based on the data stored in this memory, and a memory 24 for storing the restored data. , Another DSP 25 that performs an inverse orthogonal transform on the data stored in this memory
And an output buffer 26 for temporarily storing the digital signal which has been subjected to the inverse orthogonal transform, for transmission, are connected by a common bus 27.

The operation of the data compression / decompression device configured as described above will be described below with reference to FIG. First, in the data compression apparatus 10, the input digital signal is stored in the memory 12 via the input buffer 11. And
Of the data stored in the memory 12, a predetermined amount of data, for example, 1024 samples are subjected to orthogonal transformation to obtain frequency components, and the real part and the imaginary part are stored in the memory 14, respectively. Therefore, the DSP 15 uses the memory 14
The data stored in (1) is divided into a plurality of bands for each of the real number part and the imaginary number part, for example, eight bands, and the maximum absolute value S _m of the frequency components of each divided band is detected. In this case, the band division is equally divided on the frequency axis graduated in logarithm. Moreover, DSP 15 is greater than the maximum value S _m, and obtains the nearest power of two value SL to the maximum value S _m.

If the data width of a band is 20 bits (2 ²⁰ ), but the power SL is 2 ¹⁵ , the upper 5 bits of the frequency component data of this band are all "0". It is. Therefore, for the frequency component data in this band, the upper 5 bits can be removed (compressed) for transmission or writing. So DSP
15 is an exponent value SL after obtaining the exponent value SL, the absolute value data from which the upper bits are removed, the sign (+-) bit,
The removed bit number information, that is, the normalization scale data is sent to the encoding circuit 16. Encoding circuit 16
Sends data as a data sequence by adding synchronization data and auxiliary data such as emphasis information.

Next, the decoding circuit 21 of the data decompression device 20 extracts the absolute value data from the sent data series,
The sign bit and the normalization scale data are extracted and stored in the memory 2
Store in 2. The DSP 23 restores the real number part and the imaginary number part of the frequency component for each divided band based on the normalized scale data and stores them in the memory 24. The DSP 25 performs inverse orthogonal transformation on the frequency components stored in the memory 24 to restore the input digital signal. The restored time domain data is output via the output buffer 26.

FIG. 2 is an explanatory diagram showing the relationship between the data series input to the data compression apparatus 10 and the data series output from the data compression apparatus 10. That is, 20
When the data of 1024 samples expressed in bits is input, the total number of bits of these data is “20480”
Becomes If these data are subjected to orthogonal transformation and the higher order data bits are removed from the real number part 512 samples and the imaginary number part 512 samples of the frequency component, even if the removed bit number information is added to this, It will be 2032 "bytes or less. Even if "4" bytes of synchronization data, "4" bytes of auxiliary data, and "8" bytes of scale data are added to these data strings, the total becomes "2048" bytes. This is equivalent to compressing 20-bit data of 1024 samples into 16-bit data of 1024 samples.

By the way, depending on the voice, even if the normalization scale is determined for each divided band, the scale may continue to be large, and the amount of compressed data may be significantly larger than expected. is there. On the other hand, DSP1
5, a function for determining whether or not the data amount exceeds a predetermined value, and a component that is less necessary until the data amount falls within the predetermined value, that is,
The problem can be solved by adding a function of sequentially setting the highest frequency component to zero.

In the above embodiment, the normalization scale of the real part and the imaginary part of the frequency component is determined for each divided band, but it is known that the digital music signal does not include a special instrument sound or the like. In this case, it is not necessary to determine a normalization scale for each divided band, and encoding can be performed using a coding scale that is set to be shorter or shorter in advance, which can shorten the processing time. .

Although the above embodiment has been described with respect to the case of compressing and expanding a music signal, the present invention is not limited to this and can be applied to the compression and expansion of general audio signals.

[0032]

As described above, according to the present invention, it is possible to efficiently use a transmission system or a recording medium for a digital audio signal, especially a digital music signal, and a high-quality digital audio signal. Can be restored to. Particularly, in the data compression / decompression method according to claim 1, when the normalization scale data, the absolute value of the normalized data, and the sign bit thereof are transmitted or recorded, the receiving or reproducing side frequency component based on these data. Since it is restored to the real part and the imaginary part of the above, and the obtained frequency component is subjected to inverse orthogonal transform to be restored to a digital voice signal, it is possible to restore to a high-quality digital voice signal. The data compression / decompression method according to the second aspect deals with a case where data having a large normalization scale is continuous. Further, claim 3
The data compression / decompression method described in 1) is effective when no special musical instrument sound is included. Further, in the data compression / decompression method according to claim 4, in consideration of the symmetry of the frequency components, one of the symmetric components is removed and the absolute value data is transmitted or recorded, thereby transmitting / receiving data or recording / reproducing. The data can be reduced to approximately half, and the transmission system or the recording medium can be used more efficiently.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a data compression / decompression device embodying the present invention.

FIG. 2 is a data sequence diagram for explaining the operation of the data compression / decompression device embodying the present invention.

FIG. 3 is an explanatory diagram for explaining a schematic relationship between sample data on a time axis and sample data on a frequency axis in order to explain the principle of the present invention.

FIG. 4 is an explanatory diagram illustrating a detailed relationship between time-axis sample data and frequency-axis sample data in order to explain the principle of the present invention.

[Explanation of symbols]

 11 Input Buffer 12, 14, 22, 24 Memory 13, 15, 23, 25 Digital Signal Processor 16 Encoding Circuit 21 Decoding Circuit 27 Output Buffer

Claims (4)

[Claims] 1. A data compression / decompression method for compressing a digital audio signal on the transmitting or recording side and expanding the digital audio signal on the receiving or reproducing side to restore the audio signal, wherein the transmitting or recording side orthogonally transforms the digital audio signal. Then, the real part and the imaginary part of each obtained frequency component are divided into a plurality of bands, and the power of 2 that is larger than the maximum value of the absolute value of the frequency component and is closest to the maximum value is divided for each divided band. Determine the normalization scale, normalize the real and imaginary parts of each frequency component according to this scale, send or record the normalization scale data, the absolute value of the normalized data and its sign bit, receive Alternatively, the reproducing side restores to the real part and the imaginary part of each frequency component based on the normalized scale data, the absolute value of the normalized data and the sign bit thereof, and obtains the obtained frequency component. Data compression, decompression method performs an inverse orthogonal transform to restore the digital audio signal, characterized in that the. 2. The data compression / decompression method according to claim 1, wherein when the data amount corresponding to a predetermined number of samples exceeds a predetermined value, the highest frequency component is sequentially transmitted or recorded as zero until the predetermined value is reached. . 3. A data compression / decompression method for compressing a digital audio signal on the transmitting or recording side and expanding it on the receiving or reproducing side to restore the audio signal, wherein the transmitting or recording side orthogonally transforms the digital audio signal. Then, the real part and the imaginary part of each obtained frequency component are divided into a plurality of bands, respectively, and are normalized by a predetermined normalization scale for each divided band, and the normalization scale data or the predetermined normalization scale is used. A flag indicating that it is a scale, the absolute value of the normalized data and its sign bit are transmitted or recorded, and the receiving or reproducing side indicates that the scale data for normalization or a predetermined normalization scale is used. A digital voice by restoring the real number part and the imaginary number part of each frequency component based on the flag shown, the normalized absolute value of the data and its sign bit, and applying the inverse orthogonal transform to the obtained frequency component. Data compression, decompression method characterized by obtaining a degree. 4. The transmitting or recording side removes a frequency component of one frequency region which is symmetric with respect to a predetermined frequency to transmit or record, and the receiving or reproducing side has a symmetry of frequency components. 4. The data compression / decompression method according to claim 3, wherein the frequency components of the removed frequency domain are also restored by using the.