JP3304717B2 - Digital signal compression method and apparatus - Google Patents

Abstract

PURPOSE: To easily realize the operation of the characteristic on a frequency axis for an input signal without the need for a new mechanism and device other than the mechanism and the device to compress information without giving hindrance to cancellation condition of a reflected noise. CONSTITUTION: The device is provided with band division filters 201, 202 for QMF to divide an input signal into plural bands at information compression, MDCT circuits 203-205 to obtain data on a frequency axis of an input signal, and frequency characteristic operation circuits 219-221 to operate the characteristic on the frequency axis. The frequency characteristic operation circuits 219-221 operate the characteristic on the frequency axis so as to keep cancellation condition of reflected noise by the division and information signal expansion corresponding to the division through the band synthesis.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for recording and reproducing compressed data obtained by bit-compressing a digital audio signal and the like and a system for transmitting the compressed data. The present invention relates to a digital signal compression method and apparatus for compressing information of a digital signal by changing the frequency size of a small block subdivided according to time and frequency at which bit allocation for floating and / or compression is performed.

[0002]

2. Description of the Related Art The applicant of the present invention has previously described a technique for compressing an input digital audio signal into bits and recording the digital audio signal in a burst with a predetermined data amount as a recording unit, for example, as described in US Pat. No. 5,243,588. It has been proposed in books and drawings.

This technology uses a magneto-optical disk as a recording medium and records and reproduces AD (adaptive difference) PCM audio data defined in an audio data format of a so-called CD-I (CD-interactive) or CD-ROM XA. The ADPCM data is recorded in a burst on the magneto-optical disk using, for example, 32 sectors of ADPCM data and several sectors for linking for interleave processing as a recording unit.

Several modes can be selected for ADPCM audio in a recording / reproducing apparatus using this magneto-optical disk. For example, the sampling frequency is 37.times. At twice the compression ratio as compared to the normal CD playback time.
A level A of 8 kHz, a level B of a sampling frequency of 37.8 kHz with a 4 times compression ratio, and a level C of a sampling frequency of 18.9 kHz with an 8 times compression ratio are defined. That is, for example, in the case of the level B, the digital audio data is compressed to approximately 1/4, and the reproduction time (play time) of the disc recorded in this level B mode is a standard CD format (CD). -D
A format). Since a recording and reproducing time of the same order as a standard 12 cm can be obtained with a smaller disk, the size of the apparatus can be reduced.

However, the rotation speed of the disk is a standard C
Since it is the same as D, for example, in the case of the level B, compressed data for a reproduction time four times as long as the predetermined time is obtained. For this reason, for example, the same compressed data is read out four times in units of time such as sectors or clusters, and only one of the compressed data is used for audio reproduction. Specifically, when scanning (tracking) a spiral recording track, 1
A track jump is performed to return to the original track position for each rotation, and the reproduction operation proceeds in such a form that the same track is repeatedly tracked four times. This means that normal compressed data need only be obtained at least once out of, for example, four times of redundant reading, and is resistant to errors due to disturbances and the like, and is particularly preferable when applied to portable small devices.

[0006]

When digital audio data is compressed by applying the above-mentioned technology, in order to realize efficient compression, an auditory characteristic described later, a so-called masking effect and a minimum audible characteristic, will be described. It is effective to use such as. To take advantage of this auditory property,
It is necessary to analyze the input signal separately for each frequency,
For this purpose, orthogonal transforms and / or split filters are used. The orthogonally transformed spectrum data on the frequency axis and the data divided into at least two bands by the division filter can be handled as data on the frequency axis. Therefore, when used for changing characteristics on the frequency axis or for displaying power by frequency, that is, for displaying a so-called spectrum analyzer, there is an advantage that it can be easily configured without adding any device or system.

[0007] In the following description, the spectrum data refers to data on the frequency axis that has been orthogonally transformed by MDCT (Modified Discrete Cosine Transform), but is not necessarily limited to MDCT. That is, data including spectral components separated by a band-pass filter such as a fast Fourier transform (FFT), a discrete Fourier transform (DFT), a discrete cosine transform (DCT), and a quadrature mirror filter (QMF). Are collectively simply referred to as spectral data.

In the method of obtaining data on the frequency axis using the above-mentioned division filter, it is common to divide the band and thin out the data to prevent an increase in the number of data.
By decimating this data, aliasing noise occurs because the divided filter does not have ideal characteristics.However, this aliasing noise must be set so that the synthesis filter when decompressing the compressed data satisfies the aliasing noise cancellation condition. Is usually not a problem.

However, if the data on the frequency axis is directly used to manipulate the characteristics on the frequency axis as described above, the condition for canceling the aliasing noise will not be satisfied. For this reason, aliasing noise remains in the expanded music data, which may cause a problem in hearing, that is, a cause of deterioration of sound quality. Here, as described above, the root cause of the aliasing noise is due to imperfection of the division or synthesis filter. Therefore, by increasing the order of this filter, it is possible to bring the aliasing closer to the characteristic of the ideal filter. It is possible to eliminate the effects of noise. However, the increase in the amount of calculation in this case becomes enormous, and often does not become a practical scale.

Therefore, the present invention has been made in view of such circumstances, and divides an input signal into respective bands using a band division filter at the time of information compression, and performs band synthesis using a synthesis filter at the time of information expansion. In the device and the information compression method to be performed, when directly operating the data on the frequency axis of the input signal to give a change in the characteristics on the frequency axis, the characteristics on the frequency axis must be set so as not to disturb the aliasing noise cancellation condition. It is an object of the present invention to provide a digital signal compression method and apparatus to which a method of operating is applied.

[0011]

SUMMARY OF THE INVENTION The present invention has been proposed to achieve the above-mentioned object. A digital signal compression method and apparatus for compressing information of a digital signal according to the present invention divides an input signal into at least two bands by an analysis filter during information compression, obtains spectrum data on a frequency axis of the divided signal, It is characterized in that at least one characteristic of the on-axis spectrum data is operated based on the frequency characteristic operation information.

More specifically, when operating the characteristics on the frequency axis, the size of the spectrum data on the frequency axis is calculated, and the frequency characteristic operation information is assigned to each of the spectrum data to normalize the data. And whether or not the result of the multiplication exceeds a predetermined upper limit by using the spectrum data and the coefficient for the multiplication. Then, based on the output of this inspection, when the number of spectrum data on the frequency axis at which overflow occurs is equal to or more than a predetermined value, normalization is performed on all the coefficients for the multiplication, and this output is used as the frequency characteristic operation information. Based on this, by modifying the characteristic that maintains symmetry with the cutoff frequency of the analysis filter as the axis of symmetry, the aliasing noise is reduced so that the aliasing noise is canceled by the band synthesis by the synthesis filter, and the above correction is performed. The received coefficient is multiplied by the spectrum data on the frequency axis.

In the digital signal compression method and apparatus according to the present invention, when the characteristic on the frequency axis is manipulated,
A predetermined characteristic is obtained by changing the characteristics on the frequency axis and the condition for canceling the aliasing noise. Also, in the high frequency range, the operation of the characteristics on the frequency axis is prioritized over the condition for canceling the aliasing noise.

In the method and apparatus of the present invention, the above-described compressed data can be transmitted.

That is, the digital signal compression method and the digital signal compression apparatus according to the present invention are designed to reduce aliasing noise required for band division and synthesis filter configuration when operating the characteristics of the input signal on the frequency axis. The above-mentioned problem is solved by manipulating the characteristics of the input signal on the frequency axis so as to maintain the canceling condition.

When the characteristics on the frequency axis of the input signal are manipulated, the characteristics on the frequency axis of the input signal are manipulated so as to be symmetric with respect to the band division and the division frequency of the synthesis filter. The above-mentioned problem is solved while maintaining the condition for canceling the aliasing noise.

Furthermore, it is more effective to change the priority of the operation on the frequency axis of the input signal and the condition for canceling the aliasing noise according to the input signal and / or the application example.

On the other hand, in a general music signal, it is practical to give priority to the operation of the characteristic on the frequency axis over the cancellation of the aliasing noise in a high frequency band having relatively little energy.

That is, according to the digital signal compression method and apparatus of the present invention, in particular, a method and apparatus for dividing an input signal into respective bands using an analysis filter during information compression and performing band synthesis using a synthesis filter during information expansion. For example, when directly manipulating the spectrum data on the frequency axis of the input signal to give a change in the characteristics on the frequency axis, it is possible to manipulate the characteristics on the frequency axis so that the conditions for canceling the aliasing noise are not disturbed. This makes it possible to easily operate the characteristics on the frequency axis without requiring any new mechanism or device other than for compressing information.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

First, FIG. 1 shows a schematic configuration of a configuration example in which the present invention is applied to a compressed data recording / reproducing device as a digital signal compression device.

In the compressed data recording / reproducing apparatus shown in FIG. 1, a magneto-optical disk 1 rotated and driven by a spindle motor 51 is used as a recording medium. When recording data on the magneto-optical disk 1, for example, a so-called magnetic field modulation recording is performed by applying a modulation magnetic field corresponding to the recording data with the magnetic head 54 while irradiating the laser light with the optical head 53. The data is recorded along the recording track of. At the time of reproduction, a recording track of the magneto-optical disk 1 is traced by a laser beam by the optical head 53, and reproduction is performed magneto-optically.

The optical head 53 includes, for example, a laser light source such as a laser diode, a collimator lens, an objective lens,
The optical system includes optical components such as a polarizing beam splitter and a cylindrical lens, and a photodetector having a light receiving portion having a predetermined pattern. The optical head 53 is provided at a position facing the magnetic head 54 via the magneto-optical disk 1. When data is recorded on the magneto-optical disk 1, the magnetic head 54 is driven by a recording-system head drive circuit 66, which will be described later, to apply a modulation magnetic field in accordance with the recorded data. By irradiating the track with laser light, thermomagnetic recording is performed by a magnetic field modulation method. The optical head 53 detects reflected light of the laser beam applied to the target track, detects a focus error by, for example, a so-called astigmatism method, and detects a tracking error by, for example, a so-called push-pull method. When reproducing data from the magneto-optical disk 1, the optical head 53 detects the focus error and the tracking error, and at the same time, detects the difference in the polarization angle (Kerr rotation angle) of the reflected light of the laser light from the target track and reproduces the data. Generate a signal.

The output of the optical head 53 is supplied to an RF circuit 55. The RF circuit 55 extracts the focus error signal and the tracking error signal from the output of the optical head 53 and supplies the focus error signal and the tracking error signal to the servo control circuit 56.
Feed to 1.

The servo control circuit 56 comprises, for example, a focus servo control circuit, a tracking servo control circuit, a spindle motor servo control circuit, a thread servo control circuit and the like. The focus servo control circuit performs focus control of the optical system of the optical head 53 so that the focus error signal becomes zero. Further, the tracking servo control circuit performs tracking control of the optical system of the optical head 53 so that the tracking error signal becomes zero. Further, the spindle motor servo control circuit controls the spindle motor 51 so as to rotate the magneto-optical disk 1 at a predetermined rotation speed (for example, a constant linear speed). The thread servo control circuit moves the optical head 53 and the magnetic head 54 to target track positions of the magneto-optical disk 1 specified by the system controller 57. The servo control circuit 56 that performs such various control operations sends information indicating the operation state of each unit controlled by the servo control circuit 56 to the system controller 57.

A key input operation unit 58 and a display unit 59 are connected to the system controller 57. The system controller 57 controls a recording system and a reproduction system in an operation mode specified by operation input information from the key input operation unit 58. The system controller 57 also controls the optical head 53 and the magnetic head 5 based on the address information in sector units reproduced from the recording track of the magneto-optical disk 1 by the header time, the subcode Q data, and the like.
4 manages a recording position and a reproduction position on the recording track which are being traced. Further, the system controller 57
Controls the display unit 59 to display the reproduction time based on the data compression ratio and the reproduction position information on the recording track.

This reproduction time display is based on the reciprocal of the data compression ratio (absolute time information) with respect to sector-based address information (absolute time information) reproduced from a recording track of the magneto-optical disk 1 by so-called header time or so-called subcode Q data. For example, in the case of 1/4 compression, the actual time information is obtained by multiplying by 4), and this is displayed on the display unit 59. At the time of recording, if absolute time information is recorded in advance on a recording track of a magneto-optical disk or the like (preformatted), the preformatted absolute time information is read and the data compression ratio is adjusted. By multiplying the reciprocal, the current position can be displayed by the actual recording time.

Next, in the recording system of the data recording / reproducing apparatus, the analog audio input signal AIN from the input terminal 60 is supplied to the A / D converter 62 via the low-pass filter 61.
The A / D converter 62 quantizes the analog audio input signal AIN. The digital audio signal obtained from the A / D converter 62 is ATC (Adap
tive Transform Coding) is supplied to the PCM encoder 63. The digital audio input signal DIN from the input terminal 67 is supplied to the ATC encoder 63 via the digital input interface circuit 68. AT
The C encoder 63 performs a bit compression (data compression) process on the digital audio PCM data of a predetermined transfer rate obtained by quantizing the input signal AIN by the A / D converter 62. Here, the compression ratio will be described as 4 times, but this configuration example does not depend on this magnification, and can be arbitrarily selected depending on the application example.

Next, writing and reading of data in the memory 64 are controlled by the system controller 57.
ATC data supplied from the ATC encoder 63 is temporarily stored, and is used as a buffer memory for recording on a disk as needed. That is, for example, the compressed audio data supplied from the ATC encoder 63 has a standard C
Data transfer rate of D-DA format (75 sectors /
Second), ie, 18.75 sectors / second, and the compressed data is continuously written to the memory 64. As described above, it is sufficient for this compressed data (ATC data) to record one sector per four sectors, but such recording every four sectors is practically impossible, so that a continuous sector as described later is used. I try to keep a record. This recording is performed in bursts at the same data transfer rate (75 sectors / second) as a standard CD-DA format by using a cluster composed of a predetermined plurality of sectors (for example, 32 sectors + several sectors) as a recording unit through a pause period. It is done on a regular basis. That is, in the memory 64, ATC audio data continuously written at a low transfer rate of 18.75 (= 75/4) sectors / second corresponding to the bit compression rate is used as recording data. It is read out in bursts at the transfer speed. The overall data transfer speed of the read and recorded data, including the recording pause period, is as low as 18.75 sectors / sec. The instantaneous data transfer rate in the above is the standard 75 sectors / second. Therefore, when the disk rotation speed is the same speed (constant linear speed) as the standard CD-DA format, the same recording density and storage pattern as in the CD-DA format are recorded.

The ATC audio data, ie, recorded data, read from the memory 64 in a burst at the (instantaneous) transfer rate of 75 sectors / second is transferred to the encoder 65.
Supplied to Here, from the memory 64 to the encoder 65
In the data sequence supplied to the cluster, the unit continuously recorded in one recording is a cluster including a plurality of sectors (for example, 32 sectors) and several sectors for cluster connection arranged before and after the cluster. The cluster connection sector is set to be longer than the interleave length in the encoder 65, so that even if interleaved, data in other clusters is not affected.

The encoder 65 operates on the recording data supplied in bursts from the memory 64 as described above.
Encoding processing (parity addition and interleaving processing) for error correction, EFM encoding processing, and the like are performed. The recording data that has been subjected to the encoding process by the encoder 65 is supplied to the magnetic head drive circuit 66. The magnetic head drive circuit 66 is connected to the magnetic head 54,
The magnetic head 54 is driven so that a modulation magnetic field corresponding to the recording data is applied to the magneto-optical disk 1.

The system controller 57 controls the memory 64 as described above,
By this memory control, the recording position is controlled so that the recording data read out from the memory 64 in a burst manner is continuously recorded on the recording track of the magneto-optical disk 1.
The recording position is controlled by controlling the recording position of the recording data read in a burst from the memory 64 by the system controller 57 and transmitting a control signal designating the recording position on the recording track of the magneto-optical disk 1 to a servo control circuit. 56.

Next, a reproducing system of the data recording / reproducing apparatus will be described. This reproducing system is for reproducing the recording data continuously recorded on the recording tracks of the magneto-optical disk 1 by the above-mentioned recording system.
3, a reproduction output obtained by tracing the recording track of the magneto-optical disk 1 with a laser beam by the RF circuit 55 is provided with a decoder 71 which is binarized and supplied. At this time, not only the magneto-optical disk but also the same read-only optical disk as a compact disk (CD: COMPACT DISC) can be read.

The decoder 71 corresponds to the encoder 65 in the above-described recording system, and performs the above-described decoding processing for error correction and EFM decoding on the reproduced output binarized by the RF circuit 55. Processing such as processing is performed to reproduce audio data at a transfer rate of 75 sectors / sec, which is faster than the normal transfer rate. This decoder 7
1 is supplied to the memory 72.

In the memory 72, data writing and reading are controlled by the system controller 57, and reproduced data supplied from the decoder 71 at a transfer rate of 75 sectors / second is written in burst at the transfer rate of 75 sectors / second. It is. Also, in the memory 72, the reproduced data written in a burst at the transfer rate of 75 sectors / second is a regular transfer rate of 18.75 of 75 sectors / second.
It is read continuously at sectors / second.

The system controller 57 writes the reproduced data to the memory 72 at a transfer rate of 75 sectors / sec.
Memory control is performed such that data is read continuously at a transfer rate of 5 sectors / second. Further, the system controller 57 performs the above-described memory control for the memory 72, and reproduces the reproduction data written in a burst from the memory 72 by the memory control from the recording track of the magneto-optical disk 1 continuously. Control the playback position. This playback position control is performed by the system controller 5.
7 manages the reproduction position of the reproduction data read from the memory 72 in a burst manner,
This is performed by supplying a control signal for designating a reproduction position on a recording track of an (or an optical disk) to the servo control circuit 56.

ATC audio data obtained as reproduction data continuously read from the memory 72 at a transfer rate of 18.75 sectors / sec is supplied to an ATC decoder 73. The ATC decoder 73 reproduces 16-bit digital audio data by expanding the ATC data by four times (bit expansion). This ATC
The digital audio data from the decoder 73 is D
/ A converter 74.

The D / A converter 74 includes an ATC decoder 73
Is converted into an analog signal to form an analog audio output signal AOUT. The analog audio signal AOUT obtained by the D / A converter 74 is output from an output terminal 76 via a low-pass filter 75.

Next, in the configuration example apparatus of FIG.
The high efficiency compression encoding / decoding method or the high efficiency compression encoding / decoding apparatus of the ATC encoder 63 and the ATC decoder according to the present invention will be described in detail. That is, refer to FIG. 2 et seq. For a technique of encoding an input digital signal such as an audio PCM signal using band division coding (SBC), adaptive conversion coding (ATC), and adaptive bit allocation. I will explain while.

FIG. 2 shows a specific configuration of a high-efficiency coding apparatus for performing data compression built in the apparatus of this configuration example. In this high-efficiency coding apparatus, the input digital signal is divided into at least two frequency bands, and the bandwidths of the two lowest bands are the same, and the higher the frequency band, the higher the frequency band becomes. , Digital signals for each of these frequency bands are divided into blocks, and orthogonal transform is performed for each of these blocks, and the spectrum data on the frequency axis obtained is taken into account in the low band, taking into account the human hearing characteristics described later. Bits are adaptively allocated and coded for each so-called critical bandwidth (critical band) and for each of the subdivided critical bandwidths in consideration of the block floating efficiency in the middle and high frequency bands. Usually, this block is a quantization noise generating block. The critical band is a frequency band divided in consideration of human auditory characteristics, and a band of a pure tone when the pure tone is masked by a narrow band noise near the frequency of the pure tone. That is. In this critical band, the higher the band, the wider the bandwidth.
The entire frequency band of 2 kHz is divided into, for example, 25 critical bands.

That is, in FIG.
For example, when the sampling frequency is 44.1 kHz,
An audio PCM signal of 0 to 22 kHz is supplied. This input signal is supplied to a band dividing filter 201 such as a so-called QMF filter, for example, from 0 to 11 k.
Hz band and 11 kHz to 22 kHz band,
A signal in the 0 to 11 kHz band is also subjected to a 0 to 5.5 kHz signal by a band division filter 202 such as a so-called QMF filter.
It is divided into a Hz band and a 5.5 kHz to 11 kHz band. 11 kHz to 22 k from the band division filter 201
The signal in the Hz band is sent to the MDCT circuit 203 which is an example of the orthogonal transform circuit,
The signal in the band of 5 kHz to 11 kHz is applied to the MDCT circuit 204.
From the band division filter 202 to 0 to 5.5 k
The signals in the Hz range are sent to the MDCT circuit 205 to be subjected to MDCT processing.

Here, a QMF filter as an example of a method of dividing the input digital signal into at least two frequency bands is described in, for example, the document “Digital Coding of Speech in Subvans” (“Digi
tal coding of speech in subbands "RECrochiere,
Bell Syst. Tech. J., Vol. 55, No. 8 1976). This QMF filter has a bandwidth equal to 2 bandwidths.
This filter is characterized in that so-called aliasing does not occur when the divided bands are combined later.

Also, a document "Polyphase Quadrature Filters-New Band Division Coding Technology"("Po
lyphase Quadrature filters -A new subband coding t
echnique ", Joseph H. Rothweiler ICASSP 83, BOSTON)
Describes an equal bandwidth filter division technique.
This polyphase quadrature filter is characterized in that a signal can be divided at a time when divided into at least two bands of equal bandwidth.

Further, as the above-mentioned orthogonal transform, for example, an input audio signal is divided into blocks in a predetermined unit time (frame), and a fast Fourier transform (F
FT), a discrete cosine transform (DCT), a modified DCT transform (MDCT), and the like, there are orthogonal transforms that transform the time axis into the frequency axis. This MDCT is described in the literature "Subband / Filter Using Filter Bank Design Based on Time Domain Aliasing Cancellation."
Transform Coding "(" Subband / Transform Coding Using Filte
r Bank Designs Based on Time Domain Aliasing Cance
llation, "JPPrincen ABBradley, Univ. of Surrey
Royal Melbourne Inst. Of Tech. ICASSP 1987).

FIG. 3 shows a specific example of a standard input signal for a processing block for each band supplied to each of the MDCT circuits 203 to 205. In the specific example of FIG. 3, the three filter output signals have at least two orthogonal transform block sizes independently for each band, and the time resolution can be switched according to the time characteristic, frequency distribution, and the like of the signal. I have. If the signal is quasi-stationary in time, the orthogonal transform block size is 11.6.
ms, that is, the long mode (Lon) shown in FIG.
g Mode), and when the signal is non-stationary, the orthogonal transform block size is further divided into two and four. The short mode (Short mode) shown in FIG.
Mode), a case where all are divided into four to make 2.9 ms, or a middle mode A (Mi) shown in FIG.
As in the case of Ddle Mode A) and the middle mode B (Middle Mode B) shown in FIG. 3D, the time resolution of a part is divided into 5.8 ms in two and one part is divided in 2.9 ms. , Adapted to the actual complex input signal. It is clear that the division of the orthogonal transform block size is more effective if more complicated division is performed if the scale of the processing device allows. The block size (processing block length) is determined by the block size determination circuits 206 to 208 in FIG.
03 to 205 and output from output terminals 216 to 218 as block size information of each block.

Next, FIG. 4 is a block circuit diagram showing a schematic configuration of a specific example of the block size determining circuit. Here, the block determination circuit 206 of FIG. 2 will be described as an example. The output of 11 kHz to 22 kHz among the outputs of the band division filter 201 such as the QMF in FIG. 2 is sent to the power calculation circuit 404 via the input terminal 401 in FIG. Further, among the outputs of the band division filter 202 in FIG.
The output of 5 kHz to 11 kHz is supplied to the power calculation circuit 405 via the input terminal 402 of FIG. 4, and the output of 0 to 5.5 kHz is supplied to the power calculation circuit 40 via the input terminal 403 of FIG.
6 respectively. Further, the block size determination circuits 207 and 208 in FIG.
403 is a block size determination circuit 2
The operation is the same except for the case of 06. Each of the input terminals 401 to 403 in each of the block size determination circuits 206 to 208 has a matrix configuration, that is, the input terminal 4 of the block size determination circuit 207.
01 is 5.5 kHz of the band division filter 202 of FIG.
-11kHz output is connected and the input terminal 40
2 is connected to an output of 0 to 5.5 kHz. The same applies to the block size determination circuit 208.

In FIG. 4, each of the power calculation circuits 404 to
Reference numeral 406 determines the power of each frequency band by integrating the input time waveform for a certain period of time. At this time, the time width for integration needs to be equal to or smaller than the minimum time block among the orthogonal transform block sizes described above. Also,
Other than the above calculation method, a similar effect can be obtained by using, for example, the absolute value or the average value of the maximum amplitude within the minimum time width of the orthogonal transform block size as the representative power. The output of the power calculation circuit 404 is supplied to the change extraction circuit 408 and the power comparison circuit 409, and the power calculation circuits 405 and 406 are provided.
Are sent to the power comparison circuit 409, respectively. The change extraction circuit 408 obtains the differential coefficient of the power sent from the power calculation circuit 404 and obtains the power change information as
It is sent to the block size primary determination circuit 410 and the memory 407.

In the memory 407, a change extraction circuit 408
The transmitted power change information is stored for the maximum time of the orthogonal transform block size or more. This is because the temporally adjacent orthogonal transform blocks mutually influence each other by window processing at the time of orthogonal transform, so that the power change information of the immediately preceding temporally adjacent block is determined by the block size primary decision circuit 410. It is necessary. In the block size primary determination circuit 410, the change extraction circuit 4
08 based on the power change information of the block transmitted from the memory block 08 and the power change information of the immediately preceding block of the block adjacent to the block transmitted from the memory 407. The orthogonal transform block size of the corresponding frequency band is determined from the displacement. In this case, if a displacement equal to or more than a certain value is recognized, a shorter time orthogonal transform block size is selected. However, the effect can be obtained even if the displacement point (boundary value) is fixed. Furthermore, when the value is proportional to the frequency, that is, when the frequency is high, a large displacement causes a short block size in time, and when the frequency is low, the block size is determined to be short in time with a small displacement compared to that in the high case. , Is more effective. This value (boundary value) desirably changes smoothly, but may be a stepwise change in a plurality of stages.
The block size determined as described above is transmitted to the block size correction circuit 411.

On the other hand, in the power comparing circuit 409, the power information of each frequency band sent from each of the power calculating circuits 404 to 406 is compared at the same time and on the time axis with the time width at which the masking effect occurs, and the power is calculated. Circuit 404
The influence of another frequency band on the output frequency band is calculated and transmitted to the block size correction circuit 411. In the block size correction circuit 411, the masking information sent from the power comparison circuit 409 and the delay circuits 412 to 41
Based on the past block size information sent from each tap of the delay group consisting of 4 delays, the block size sent from the block size primary determination circuit 410 is modified to select a longer block size, The signal is output to the delay circuit 412 and the window shape determination circuit 415. The operation of the block size correction circuit 411 is as follows.
Even if the pre-echo becomes a problem in the frequency band concerned, in other frequency bands, especially in a band lower than the frequency band concerned, if there is a signal having a large amplitude, due to the masking effect, the pre-echo does not cause a problem in hearing, or It takes advantage of the property that problems may be reduced. The masking refers to a phenomenon in which a certain signal masks another signal and becomes inaudible due to human auditory characteristics. The masking effect includes time-axis masking by an audio signal on a time axis. There is an effect and a simultaneous masking effect by a signal on the frequency axis. Due to these masking effects, even if there is noise in the masked part,
This noise will not be heard. For this reason, in an actual audio signal, noise within the masked range is regarded as acceptable noise.

Next, in the delay group composed of the delay circuits 412 to 414, the past orthogonal transform block sizes are sequentially recorded, and each tap, that is, the delay circuits 412 to 414 is recorded.
From the output of 414, it is output to the block size determination circuit 411. At the same time, the output of the delay circuit 412 is connected to the output terminal 417, and the outputs of the delay circuits 412 and 413 are connected to the window shape determination circuit 415. Outputs from the delay circuits 412 to 414 are used by the block size correction circuit 411 to use the change in block size over a longer time period to determine the block size of the corresponding block. If the size is selected, increase the selection of the temporally shorter block size, and if the temporally shorter block size has not been selected in the past,
It is possible to make a decision such as increasing the selection of a block size that is long in time. It should be noted that the number of taps of this delay group may be increased or decreased depending on the actual configuration and scale of the apparatus, except for the delay circuits 412 and 413 required for the window determination circuit 415 and the output terminal 417.

In the window shape determination circuit 415, the output of the block size correction circuit 411, that is, the size of the next block adjacent to the block in time and the output of the delay circuit 412, that is, the block size of the block and the delay circuit From the output of 413, that is, the size of the immediately preceding block of the block in question, the shape of the window used in each of the MDCT circuits 203 to 205 in FIG. 2 described above is determined and output to the output terminal 416. . The output terminal 417 of FIG. 4, that is, the block size information and the output terminal 416, that is, the window shape information,
Are connected to the respective units as outputs of the block size determination circuits 206 to 208.

Here, the window shape determined by the window shape determining circuit 415 will be described. FIG. 5 shows the shape of the adjacent block and window. As can be seen from FIGS. 5 (a) to 5 (c), the window used for orthogonal transformation has a portion overlapping with a temporally adjacent block as shown by a dotted line and a solid line in FIG. In the example, since the shape overlapping the center of the adjacent block is adopted, the shape of the window changes depending on the orthogonal transformation size of the adjacent block.

FIG. 6 shows details of the window shape. In FIG. 6, window functions f (n) and g (n +
N) is given as a function satisfying the following equation (1). f (n) × f (L-1-n) = g (n) × g (L-1-n) f (n) × f (n) + g (n) × g (n) = 1 (1) 0 ≦ n ≦ L−1.

L in the equation (1) is the conversion block length as it is if the adjacent conversion block lengths are the same, but if the adjacent conversion block lengths are different, the shorter conversion block length is set to L. If the longer conversion block length is K, in a region where the windows do not overlap, it is given by the following equation (2). f (n) = g (n) = 1 K ≦ n ≦ 3K / 2−L / 2 f (n) = g (n) = 0 3K / 2 + L ≦ n ≦ 2K (2) In this manner, the overlapping portion of the window Is taken as long as possible to improve the frequency resolution of the spectrum at the time of orthogonal transformation. As is clear from the above explanation,
The shape of the window used for the orthogonal transform is determined after the orthogonal transform size of three temporally continuous blocks is determined. Therefore, the block of the signal input from the input terminals 401 to 403 in FIG. 4 and the block of the signal output from the output terminals 416 and 417 have a difference of one block in the present configuration example.

The power calculation circuits 405 and 40 shown in FIG.
6 and the power comparison circuit 409 can be omitted to configure the block size determination circuits 206 to 208 in FIG. Further, by fixing the shape of the window to the smallest temporal block size that the orthogonal transform block can take, the type is made one type, and the delay 41 shown in FIG.
2 to 414, the block size correction circuit 411, and the window shape determination circuit 415 may be omitted. In particular, in an application example that does not like the delay of the processing time, the above-described omission causes a configuration with a small delay,
Works effectively.

Referring again to FIG. 2, each MDCT circuit 20
The spectrum data on the frequency axis, that is, the MDCT coefficient data obtained by performing the MDCT processing in 3 to 205, is applied to each frequency characteristic operation circuit 219 to 221 and bit allocation (allocation).
It is transmitted to the calculation circuit 209.

In each of the frequency characteristic operation circuits 219 to 221, the frequency characteristic operation information inputted from each of the input / output terminals 222 to 224, that is, the frequency characteristic operation information specified by the system controller 57 in FIG. 1 and each of the block determination circuits 206 to 208.
Each MDCT based on the block size information transmitted
The frequency characteristics of the spectrum data on the frequency axis transmitted from the circuits 203 to 205 are manipulated and transmitted to the adaptive bit allocation encoding circuits 210 to 212. Further, the signal is transmitted from the input / output terminals 222 to 224 to the system controller 57 in FIG. 1 and used for displaying power for each frequency, that is, for displaying a so-called spectrum analyzer.

FIG. 7 is a block diagram showing a schematic configuration of one specific example of each of the frequency characteristic operation circuits 219 to 221 in FIG. In FIG. 7, the frequency characteristic operation circuit 219 will be described for easy understanding, but it is apparent that each of the frequency characteristic operation circuits 220 and 221 has the same configuration.

In FIG. 7, an input terminal 301 has
Spectral data (SD) on the frequency axis is supplied from the MDTC circuit 203 in FIG. 2 and sent to the overflow inspection circuit 304 and the multiplier 309 in FIG. The input terminal 302 is supplied with block information (BI) from the block determination circuit 206 shown in FIG. 2, and the input terminal 303 is supplied with the input / output terminal 222 shown in FIG. 2, that is, from the system controller 57 shown in FIG. Frequency characteristic operation information (CTL) is supplied,
It is sent to the multiplication coefficient calculation circuit 305.

The multiplication coefficient calculation circuit 305 calculates the magnitude at each frequency of the spectrum data or MDCT coefficient data on the frequency axis based on the block information (BI), and converts the frequency characteristic operation information (CTL) to the previous value. Normalization is performed by allocating each spectrum data (SD) on the frequency axis, a coefficient for multiplication is calculated, and the coefficient is transmitted to the coefficient correction circuit 306 and the overflow inspection circuit 304. By adopting a method of directly operating the spectrum data or MDCT coefficient data on the frequency axis in this manner, a predetermined characteristic can be easily obtained only by multiplying the normalized coefficient.

Next, in the overflow inspection circuit 304, the multiplication result exceeds a predetermined upper limit by the spectrum data (SD) on the frequency axis transmitted from the input terminal 301 and the coefficient transmitted from the multiplication coefficient calculation circuit 305. Are you testing. The upper limit is defined by the word length of the input / output signal, the operation accuracy of the device, the circuit, and the like. In this configuration example, the upper limit is set to 2 64. When an overflow is confirmed in this inspection, the overflow inspection circuit 304 transmits the information to the coefficient correction circuit 306, and when no overflow occurs, transmits no signal.

In this configuration example, the following two processes are selected or used in combination for the overflow process. First, as a result of multiplication, if the number of spectrum data (SD) on the frequency axis where overflow occurs is small,
Only the coefficient of the data in which an overflow occurs is corrected so that the multiplication result is equal to or less than the upper limit of the data. This operation is close to a so-called amplitude limiting (limiter) operation on data on the time axis. Next, when the number of spectrum data (SD) on the frequency axis at which an overflow occurs is large, the entire multiplication coefficient is normalized so that the data with the maximum multiplication result does not exceed the upper limit. This operation is similar to the operation of compressing the amplitude of data on the time axis.

The modification of the multiplication coefficient for preventing the overflow is more effective if it is changed according to the characteristics of the input signal and the compression method. In the present configuration example, weighting is given according to the frequency of the peak component of the input signal, and the above-described correction of the multiplication coefficient is executed, thereby obtaining a good result.

Further, as described above, when the block corresponds to the processing block where the overflow occurs, the coefficient correction circuit 306 immediately corrects the coefficient. When the processing is performed and the overflow processing is not necessary in the processing block, the processing is performed over at least two blocks by referring to the correction state of the coefficient of the processing block temporally output from the memory 308. The control is performed so that the coefficient is corrected with a significant change. This control is more effective if it is changed according to the application or the characteristics of the input signal. The multiplication coefficient corrected as described above is transmitted to the aliasing noise reduction circuit 307.

In the aliasing noise reduction circuit 307, each frequency characteristic operation circuit 219, 220, 22 in FIG.
By operating the frequency characteristic in No. 1, the coefficient is changed so as to reduce aliasing noise that is not canceled. In the present configuration example, this aliasing noise is caused by each of the QMF filters 201 and 20 in FIG.
2 occurs when the data is divided and thinned out by using
6, the noise is canceled when the band synthesis is performed by the band synthesis filters 101 and 102 and does not normally cause a problem. However, when the characteristics on the frequency axis are changed, the above-described cancellation becomes insufficient, A problem may arise.

Here, the aliasing noise generated by the band division filter composed of the QMF will be described.
As mentioned earlier, using a filter to divide the signal into two equal-bandwidth bands, and then sampling at half the rate, sampling is halted in both bands due to inevitable filter imperfections. Aliasing occurs.
However, when QMF is used as a filter, when two signals are combined and returned to the original band signal, both aliasing noise components can be canceled.

First, in a filter composed of QMF,
The condition under which the aliasing noise component is canceled will be described. The filter of the QMF is a symmetric F filter having an even order L.
Use an IR filter (Symmetrical FIR Filter). When the impulse response of the low pass filter F ₁ which limits the half band f _{1 (n),} and to the Z-transform F ₁ and _(z), F 1 _(z) is be represented by the following formula (3) Can be done.

[0068]

(Equation 7)

From the low-pass filter of the equation (3), the following equation is obtained.
It is possible to configure the high-pass filter F ₂ having an impulse response represented by (4).

[0070]

(Equation 8)

The filters of F ₁ and F ₂ are connected to the input signal train x
, X ₁ , x expressed by the following equation (5)
You can get ₂ .

[0072]

(Equation 9)

Further, x ₁ and x ₂ are each decimated to a half rate, and y ₁ and y ₂ represented by the following equation (6), that is, a signal sequence divided into two bands of equal bandwidth are obtained. create.

[0074]

(Equation 10)

[0075] However, since the signal y ₂ of the high band side is folded back at the time of thinning it to a lower frequency, for example, the highest frequency components of x ₂ are decimated, observed as the lowest frequency signal as y ₂ Is done. In this configuration example, such a Q
After the filter is divided into three bands using two stages of MF filters, orthogonal transformation is performed and information compression is performed. Here, y ₁ and y ₂ represented by this equation (6) are used as they are. The description will be made on the assumption that the signal sequence before the division is synthesized. y ₁ ,
y ₂ is a signal sequence including an aliasing noise component, which is synthesized as follows. First, zero is interpolated into y ₁ and y ₂ to generate signal sequences u ₁ and u ₂ represented by the following equation (7).

[0076]

[Equation 11]

Next, based on the filter of the above equation (3),
Filters G ₁ and G having an impulse response represented by (8)
₂ are made to act on u ₁ and u ₂ represented by the equation (7), respectively, and t ₁ and t 2 represented by the equation (9) are formed.
Produce ₂ .

[0078]

(Equation 12)

[0079]

(Equation 13)

Assuming that s is expressed by equation (10), a relation expressed by equation (11) is established between the Z conversion X of x and the Z conversion S of s.

S (n) = t ₁ (n) + t ₂ (n) (10)

[0082]

[Equation 14]

Here, since F ₁ is a symmetrical FIR filter, it has a linear phase characteristic and can be expressed by equation (12). Using this, when L is an even number,
(13) is established.

[0084]

(Equation 15)

[0085]

(Equation 16)

[0086] Now Equation (14) sets an impulse response f ₁ of the low-pass filter F ₁ so as to hold, s is half the amplitude of the x, consistent with those obtained by L-1 sample delay.

[0087]

[Equation 17]

That is, this means that the aliasing noise components included in y ₁ and y ₂ represented by the equation (6) have disappeared. From the above, conditions for canceling the aliasing noise in the filter made of QMF is the impulse response of the low pass filter F ₁ to be used as f _{1 (n),} and the Z-transform F ₁ and _(n) In this case, the above-described equation (14) holds, and the impulse response f _{2 of the} high-pass filter F ₂ paired with the low-pass filter is
Equation (4) is satisfied. Furthermore, it is necessary that the impulse responses g ₁ and g _{2 of} the synthesis filters G ₁ and G ₂ satisfy Expression (8).

Here, using FIGS. 8A to 8H,
The aliasing noise described above will be described. FIG. 8A shows the characteristics on the frequency axis when dividing into equal bands by QMF. The low-pass filter and the high-pass filter have frequency characteristics that are symmetric with respect to the stop frequency fs / 4. In FIG. 8A centered on fs / 4, the frequency region indicated by A is:
It shows a frequency region in which aliasing noise is strongly generated due to imperfect characteristics of each filter. Here, assuming that the signal sg indicated by the up arrow in the figure is input to the filter system, the output of each filter is indicated by the up arrow in the diagrams of FIGS. 8B and 8E. A signal sg shown in FIG. Although the signal sg exists in the pass band of the high-pass filter, it also appears in the output of the low-pass filter due to imperfect filter characteristics.

In this state, decimation (De
8 (c) and (f) of FIG. When the decimation is performed, the sampling frequency becomes fs / 2. Therefore, the data of the frequency of fs / 4 or more in the low frequency range and the frequency of fs / 4 or less in the high frequency range are shown in FIG. 8C and FIG. ), So that the component is apparently seen as a component of fs / 4 or less. In the high band, the frequency component of fs / 4 or more is turned back, so that the data of the lowest band apparently indicates the data of the actual highest band. Note that FIG. 8F shows a high-frequency component.

The aliasing noise generated by the decimation appears as frequency components as shown in FIGS. 8 (d) and 8 (g) by performing interpolation for synthesis. Become. Here, by performing synthesis using a filter system that satisfies the above-described aliasing noise canceling condition, aliasing distortion generated on the low frequency side is canceled as shown in FIG. The same signal as the input shown in FIG. 8A is obtained.

Next, the state of aliasing noise when the characteristic on the frequency axis is changed as described in the present configuration example and a method of canceling the aliasing noise will be described. FIG. 9A is the same diagram as FIG. 8C, and shows the result of performing decimation by inputting the signal shown in FIG. 8A through a low-pass filter. Similarly, (d) of FIG. 9 is the same as (f) of FIG. 8, and shows the result of performing decimation by passing the signal shown in (a) of FIG. 8 through a high-pass filter. I have.

Here, the result of the decimation through the above low-pass filter, that is, the result shown in FIG. 9A is further converted to fs / 4 as shown in FIG. 9B.
When the characteristic operation on the frequency axis is performed to increase the intensity in the vicinity, the aliasing noise generated by the previous decimation is also enhanced as shown in FIG. In this state, even if interpolation is performed as shown in FIG. 9C, the aliasing noise remains emphasized.

Even if the output shown in (c) of FIG. 9 is combined with the data on the high frequency side shown in (e) of FIG. 9 as shown in (f) of FIG. Is emphasized, the aliasing noise is not canceled as shown in FIG. 9 (g).

Therefore, a characteristic change on the frequency axis after the band is divided using the QMF system as in the present configuration example results in that aliasing noise is not canceled.

Here, FIG. 10A to FIG. 10H
Thus, the aliasing noise is canceled even when the characteristic change on the frequency axis described above is given, and the method of the present invention will be described. FIG. 10A shows the result of performing decimation by inputting the same signal as in FIG. 8C and FIG. 9A through a low-pass filter. Similarly, (d) of FIG. 10 corresponds to (f) of FIG.
(D) shows the result of decimation by inputting the same signal as above and passing it through a high-pass filter.

Further, in FIGS. 10 (a) to 10 (h), FIGS. 10 (a) and 10 (b) and FIG.
(C) shows a state on the low frequency side, and is the same as (a) in FIG. 9, (b) in FIG. 9, and (c) in FIG. Here, FIG.
As shown in (e) of FIG.
When the same operation as the characteristic change on the frequency axis shown in FIG. 10B is performed to emphasize the signal components and then interpolation is performed, a data string shown in FIG. 10F is obtained. By combining the data shown in FIG. 10 (f) with the low-frequency side data shown in FIG. 10 (c) as shown in FIG. 10 (g), it is possible to cancel aliasing noise. In this case, as shown in FIG. 10 (f), the characteristic changes on the frequency axis shown in FIG. 10 (b) to FIG. And f
It becomes a line symmetry centering on s / 4. Therefore, when a change in the frequency characteristic is applied to the data on the frequency axis arranged in the order of the frequency, if a characteristic that is axisymmetric to fs / 4 is taken, the aliasing noise is canceled.

Further, FIGS. 11 (a) to 11 (d)
The operation of the above-described aliasing noise reduction circuit 307 will be described with reference to FIG. FIG. 11A is a diagram illustrating a state of band division in the present configuration example. As described above, in this configuration example, after dividing the input signal into three bands, FIG.
The so-called orthogonal transform is performed by each of the MDCT circuits 203, 204, and 205.

Here, assuming that the low-pass characteristics as shown in FIG. 11B are given in the frequency characteristic operation circuits 219, 220, and 221 in FIG. 2, the characteristics near the band division frequency fs / 4 are obtained. Greatly changes, and the symmetry between the low frequency side and the high frequency side is broken with fs / 4 as the symmetry axis on the frequency axis. As a result, the aliasing noise is not canceled at the time of the above-described band synthesis, but appears in the output. If the input signal has a certain amount of energy near fs / 4, there is a problem in the sense of hearing. There is.

Thus, the aliasing noise reduction circuit 30 shown in FIG.
In FIG. 7, as described above, as shown in FIG. 11 (c), the aliasing noise is canceled by canceling the aliasing noise by modifying the characteristic that maintains the symmetry with fs / 4 as the axis of symmetry. Have been around.

Further, if it is desired to secure a certain amount of high-frequency attenuation in accordance with the input signal, or if aliasing noise generated by the high-frequency signal is unlikely to be a problem, FIG. It is more effective if the attenuation as shown by the dashed line in FIG. 11D is made such that the attenuation in the high frequency range is kept constant or the symmetry is maintained up to a certain frequency and then attenuated. In this configuration example, good results are obtained by selecting and / or using the above-mentioned correction method in accordance with the high-frequency energy or power of the input signal and the desired attenuation in the high frequency.

Here, in FIG. 7 again, the multiplication coefficient corrected by the aliasing noise reduction circuit 307 as described above is transmitted to the multiplier 309 and the spectrum (SD ), Undergoes a predetermined characteristic operation, and is transmitted via an output terminal 310 to the adaptive bit allocation encoding circuit 210 in FIG. Further, each data subjected to the above-described characteristic operation is input / output terminals 222 and 223 in FIG. 2 for power display by frequency, so-called spectrum analyzer display and the like.
The system controller 5 in FIG.
7 and displayed on the display unit 59.

Referring again to FIG. 2, bit allocation calculating circuit 2
In step 09, the masking amount of each divided band in consideration of the critical band and the block floating is determined in consideration of the so-called masking effect, and the energy or energy of each divided band in consideration of the masking amount and the critical band and the block floating is calculated. The number of bits to be allocated is determined for each band based on the peak value and the like, and is transmitted to each of the adaptive bit allocation coding circuits 210, 211, and 212. Each adaptive bit allocation encoding circuit 210, 211, 21
In step 2, each spectrum data or MDCT coefficient data is quantized according to the number of bits allocated to each band. The data encoded in this way is taken out via the output terminals 213, 214, 215.

Next, FIG. 12 shows the bit allocation calculating circuit 2
FIG. 9 is a block circuit diagram showing a schematic configuration of a specific example of the embodiment 09. The operation of the bit allocation calculation circuit will be described with reference to FIG. 12, an input terminal 701 is connected to each of the MDCT circuits 203, 204, 20 in FIG.
Spectrum data or MDCT on frequency axis from 5
Coefficient data is supplied. The input data on the frequency axis is sent to the energy calculation circuit 702 for each band,
The energy of each divided band in consideration of the masking amount, the critical band, and the block floating is obtained, for example, by calculating the sum of the respective amplitude values in the band. Instead of the energy for each band, a peak value or an average value of the amplitude value may be used. As an output from the energy calculation circuit 702, for example, the spectrum of the sum value of each band is shown as SB in FIG. However, in FIG. 13, for simplicity of illustration, the number of divided bands in consideration of the masking amount, the critical band, and the block floating is represented by 12 bands (B _{1 to} B ₁₂ ).

Here, in order to consider the influence of the spectrum SB on so-called masking, convolution (convolution) processing is performed such that the spectrum SB is multiplied by a predetermined weighting function and added. Therefore, the output of the energy calculation circuit 702 for each band, that is, each value of the spectrum SB is sent to the convolution filter circuit 703. The convolution filter circuit 703 includes, for example, at least two delay elements for sequentially delaying input data, and at least two multipliers (for example, corresponding to each band) for multiplying an output from these delay elements by a filter coefficient (weighting function). 25 multipliers) and a sum adder for summing the outputs of the multipliers. By this convolution processing, the sum of the parts indicated by the dotted lines in FIG. 14 is obtained.

Here, a specific example of the multiplication coefficient (filter coefficient) of each multiplier of the convolution filter circuit 703 will be described. When the coefficient of the multiplier M corresponding to an arbitrary band is 1, the multiplier M-1 is a coefficient of 0.15, multiplier M-2 is a coefficient of 0.0019, and multiplier M-3 is a coefficient of 0.00000.
086, a coefficient 0.4 by a multiplier M + 1, and a multiplier M + 2
Is multiplied by the coefficient 0.06 by the multiplier M + 3 and the coefficient 0.007 by the multiplier M + 3, thereby performing the convolution processing of the spectrum SB. Here, M is an arbitrary integer of 1 to 25.

Next, the output of the convolution filter circuit 703 is sent to a subtractor 704. The subtractor 704 calculates a level α corresponding to an allowable noise level described later in the convolved area. The level α corresponding to the permissible noise level (permissible noise level) becomes an allowable noise level for each of the critical bands by performing inverse convolution processing as described later. Level. here,
The subtractor 704 is supplied with an allowance function (a function expressing a masking level) for obtaining the level α. The level α is controlled by increasing or decreasing the allowable function. The permissible function is supplied from the (n-ai) function generation circuit 705 described below.

That is, the level α corresponding to the allowable noise level can be obtained by the following equation (15), where i is a number sequentially given from the lower band of the critical band.

Α = S− (n−ai) (15) In the equation (15), n and a are constants, a> 0, and S is the intensity of the convolution-processed bark spectrum.
-Ai) is the tolerance function. In this configuration example, n = 38,
Since a = 1, there was no deterioration in sound quality at this time, and good encoding was performed.

Thus, the level α is obtained, and this data is transmitted to the divider 706. The divider 706 is for performing inverse convolution of the level α in the convolved region. Therefore, by performing this inverse convolution processing, a masking spectrum can be obtained from the level α. That is, this masking spectrum becomes an allowable noise spectrum. Note that the above inverse convolution processing requires a complicated operation, but in this configuration example, the inverse convolution is performed using a simplified divider 706.

Next, the masking spectrum is transmitted to a subtractor 708 via a synthesis circuit 707. Here, the output from the energy detection circuit 702 for each band, that is, the spectrum S
B is supplied via a delay circuit 709. Therefore, the subtractor 708 performs a subtraction operation between the masking spectrum and the spectrum SB, thereby obtaining the signal shown in FIG.
As shown, the spectrum SB is masked below the level indicated by the level of the masking spectrum MS.

The output from the subtracter 708 is taken out via an allowable noise correction circuit 710 and an output terminal 711.
M and the like (not shown). The ROM or the like assigns each band in accordance with the output (the level of the difference between the energy of each band and the output of the noise level setting means) obtained from the subtraction circuit 708 via the allowable noise correction circuit 710. Output bit number information. This information on the number of allocated bits is used for each of the adaptive bit allocation encoding circuits 210, 2 in FIG.
11 and 212, each MDC in FIG.
Each spectrum data on the frequency axis from the T circuits 203, 204, and 205 is quantized by the number of bits assigned to each band.

That is, in summary, in each of the adaptive bit allocation encoding circuits 210, 211, and 212 in FIG. 2, the energy of each divided band in consideration of the masking amount, the critical band and the block floating, and the noise level setting means The spectrum data for each band is quantized by the number of bits allocated according to the level of the difference from the output. Note that the delay circuit 709 in FIG. 12 takes the spectrum S from the energy detection circuit 702 into account in consideration of the amount of delay in each circuit before the synthesis circuit 707.
It is provided to delay B.

By the way, at the time of synthesizing by the synthesizing circuit 707, data indicating a so-called minimum audible curve RC which is a human auditory characteristic as shown in FIG. The masking spectrum MS can be synthesized. At this minimum audible curve, if the absolute noise level is below this minimum audible curve, the noise will not be heard. Although the minimum audible curve is different depending on the reproduction volume at the time of reproduction, for example, even if the coding is the same, in a realistic digital system, for example, music is input into a 16-bit dynamic range. Since there is not much difference, if quantization noise in the most audible frequency band around 4 kHz is not heard, for example, it is considered that quantization noise below the level of the minimum audible curve is not heard in other frequency bands. .

Accordingly, it is assumed that the system is used in such a manner that noise near the word length of the system, for example, around 4 kHz cannot be heard, and an allowable noise level is obtained by synthesizing the minimum audible curve RC and the masking spectrum MS together. In this case, the allowable noise level in this case can be up to the shaded portion in FIG. In this configuration example, the 4 kHz level of the minimum audible curve is adjusted to the lowest level corresponding to, for example, 20 bits. FIG. 15 also shows the signal spectrum SS.

In the allowable noise correction circuit 710,
The permissible noise level in the output from the subtracter 708 is corrected based on, for example, information on the equal loudness curve sent from the correction information output circuit 713. Here, the equal loudness curve is a characteristic curve relating to human auditory characteristics. For example, the loudness curve is obtained by calculating the sound pressure of sound at each frequency that sounds as loud as the pure tone of 1 kHz, and connecting the curves with each other. Also called a sensitivity curve. This equal loudness curve is the minimum audible curve R shown in FIG.
It draws substantially the same curve as C. In this equal loudness curve, for example, at around 4 kHz, even if the sound pressure falls by 8 to 10 dB below 1 kHz, it sounds as large as 1 kHz, and conversely, at around 50 Hz, the sound pressure must be about 15 dB higher than the sound pressure at 1 kHz. It doesn't sound the same size. For this reason, it can be seen that noise exceeding the level of the minimum audible curve (allowable noise level) should have a frequency characteristic given by a curve corresponding to the equal loudness curve. From this, it can be seen that correcting the allowable noise level in consideration of the equal loudness curve is suitable for human auditory characteristics.

Further, in the correction information output circuit 713, each of the adaptive bit allocation encoding circuits 210, 211, 212
The above-described allowable noise level is corrected based on information on an error between a detection output of an output information amount (data amount) at the time of quantization in and a target bit rate value of final encoded data. This is because the total number of bits obtained by previously performing temporary adaptive bit allocation for all the bit allocation unit blocks is a fixed number of bits (target value) determined by the bit rate of the final encoded output data. May have an error, and the bits are allocated again so that the error becomes zero. That is, when the total number of allocated bits is smaller than the target value, the difference bit number is allocated to each unit block and added, and when the total allocated bit number is larger than the target value, the difference bit number is allocated to each unit block. It works by allocating to and shaving.

In order to perform the above operation, an error of the total allocated bit number from the target value is detected, and correction information output circuit 713 corrects each allocated bit number in accordance with the error data. Output data. Here, when the error data indicates that the number of bits is insufficient, it is possible to consider a case where the data amount is larger than the target value by using a large number of bits per unit block. When the error data is data indicating the remainder of the number of bits, it can be considered that the number of bits per unit block is small and the data amount is smaller than the target value.

Therefore, the correction information output circuit 713
In accordance with the error data, the data of the correction value for correcting the allowable noise level in the output from the subtractor 708 based on, for example, the information data of the equal loudness curve is output. . By transmitting the correction value as described above to the allowable noise correction circuit 710, the allowable noise level from the subtractor 708 is corrected. In the system as described above, data obtained by processing the orthogonal transform output spectrum with sub-information as main information and a scale factor indicating a block floating state and a word length indicating a word length are obtained as sub-information, and are transmitted from the encoder to the decoder. Can be

FIG. 16 shows the ATC decoder 7 in FIG.
No. 3, that is, a decoding circuit for decoding again the signal which has been encoded with high efficiency as described above. The quantized MDCT coefficient of each band, that is, data equivalent to the output signal of each output terminal 213, 214, 215 in FIG.
The block size information used and input to the decoding circuit input 107 and information on the frequency length of the small block for block floating and quantization, that is, the output signal of each output terminal 216, 217, 218 in FIG. Is provided to the input terminal 108. The adaptive bit allocation decoding circuit 106 releases bit allocation using the adaptive bit allocation information. Next, each inverse orthogonal transform (I
MDCT) circuits 103, 104, and 105 convert signals on the frequency axis into signals on the time axis. The signals on the time axis of these partial bands are converted into respective band synthesis filters (IQMF).
The signals are decoded by the circuits 102 and 101 into full-band signals.

The present invention is not limited to the above configuration example. For example, the above-mentioned recording / reproducing medium and a signal compression device or a decompression device, or a signal compression device and a decompression device are integrated. It is not necessary to connect the data, and it is also possible to connect between them by a data transfer line, an optical cable, communication by light or radio waves, etc., without using a recording medium. Further, for example, the present invention is applicable not only to audio PCM signals but also to signal processing devices for digital audio (speech) signals and digital video signals.

Further, the recording medium records data compressed by the digital signal processing device,
Effective use of the recording capacity can be achieved. As the recording medium, not only the optical disk described above, but also a magnetic disk,
Various recording media such as an IC memory and a card incorporating the memory and a magnetic tape can also be used.

[0123]

As is clear from the above description, according to the digital signal compression method and apparatus of the present invention, the band of the input signal is divided by using the analysis filter, and the divided signal is divided on the frequency axis. A mechanism for compressing information and / or a new mechanism other than a device, etc., when disturbing aliasing noise cancellation conditions when directly changing spectrum data to give a characteristic change on the frequency axis. In addition, the operation of the characteristics on the frequency axis of the input signal can be easily realized without requiring any device.

[Brief description of the drawings]

FIG. 1 is a block circuit diagram showing a configuration example of a compressed data recording / reproducing device (disk recording / reproducing device) as a digital signal compressing device of the present invention for realizing the digital signal compressing method of the present invention.

FIG. 2 is a block circuit diagram illustrating a specific example of a high-efficiency compression-encoding encoder that can be used for bit-rate compression-encoding in the present configuration example.

FIG. 3 is a diagram illustrating a structure of an orthogonal transform block at the time of bit compression.

FIG. 4 is a block circuit diagram illustrating a configuration example of a circuit that determines an orthogonal transform block size.

FIG. 5 is a diagram illustrating a relationship between a change in temporal length of orthogonally adjacent orthogonal transform blocks and a window shape used at the time of orthogonal transform.

FIG. 6 is a diagram illustrating a detailed example of a window shape used at the time of orthogonal transformation.

FIG. 7 is a block circuit diagram illustrating a configuration example of a circuit that performs a frequency characteristic operation.

FIG. 8 is a diagram illustrating how aliasing noise is generated and canceled in a filter including a QMF.

FIG. 9 is a diagram illustrating a change in aliasing noise when a characteristic change on the frequency axis is given after band division is performed by a filter including a QMF.

FIG. 10 is a diagram showing conditions that enable a change and cancellation of aliasing noise when a characteristic change on the frequency axis is given after band division is performed by a filter including a QMF.

FIG. 11 is a diagram showing the progress of the operation of the aliasing noise reduction circuit in the present configuration example.

FIG. 12 is a block circuit diagram showing an example of a bit allocation calculation circuit using a convolution operation for realizing a bit allocation operation function.

FIG. 13 is a diagram illustrating spectra of each critical band and a band divided in consideration of block floating.

FIG. 14 is a diagram showing a masking spectrum.

FIG. 15 is a diagram in which a minimum audible curve and a masking spectrum are combined.

FIG. 16 is a block circuit diagram showing a specific example of a high efficiency compression encoding decoder that can be used for the bit rate compression encoding of the above configuration example.

[Explanation of symbols]

101, 102 Band synthesis filter (IQMF) 103, 104, 105 Inverse orthogonal transform circuit (IMDC
T) 106 adaptive bit allocation decoding circuit 201, 202 band division filter 203, 204, 205 orthogonal transform circuit (MDCT) 206, 207, 208 block determination circuit 209 bit allocation calculation circuit 210, 211, 212 adaptive bit allocation coding circuit 304 Overflow inspection circuit 305 Multiplication coefficient calculation circuit 306 Coefficient correction circuit 307 Aliasing noise reduction circuit 404, 405, 406 Power calculation circuit 408 Change extraction circuit 409 Power comparison circuit 410 Block size primary determination circuit 411 Block size correction circuit 412, 413 , 414 delay group 415 window shape decision circuit

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁷ identification code FI H04B 14/00 G10L 9/18 C (58) Fields investigated (Int.Cl. ⁷ , DB name) H03M 7/30 G10L 19 / 00 H03H 17/02

Claims (10)

(57) [Claims] 1. A digital signal compression method for compressing information of a digital signal, comprising: a dividing step of dividing an input signal into at least two bands by an analysis filter; A conversion step of obtaining data; and an operation step of operating at least one characteristic of the spectrum data on the frequency axis based on frequency characteristic operation information, wherein the operation step of operating the characteristic comprises: Calculate the size of the above spectral data,
A multiplication coefficient calculation step of allocating and normalizing the frequency characteristic operation information for each of the spectrum data and outputting a coefficient for multiplication, and a multiplication result is previously determined by the spectrum data and the coefficient supplied from the multiplication coefficient calculation step. An overflow inspection step of inspecting whether the value exceeds a predetermined upper limit; and, based on an output of the overflow inspection step, when the number of spectrum data on the frequency axis at which an overflow occurs is equal to or greater than a predetermined value, A coefficient correction step of normalizing the entire coefficient, and correcting the output of the coefficient correction step to a characteristic that maintains symmetry with the cutoff frequency of the analysis filter as a symmetric axis based on the frequency characteristic operation information,
An aliasing noise reduction step of reducing aliasing noise so as to maintain a condition for canceling aliasing noise by band synthesis by a synthesis filter; and multiplying the coefficient corrected in the aliasing noise reduction step by the spectrum data on the frequency axis. A digital signal compression method. 2. The dividing step of dividing the input signal into at least two bands includes performing band division using a quadrature mirror filter including at least a low-pass filter and a high-pass filter, and operating the characteristic. The operation step is to set the impulse response of the low-pass filter to f ₁ (n), and
₁ (n), Holds, and the impulse response f ₂ of the high-pass filter is given by Is satisfied, and the impulse responses g ₁ and g _{2 of the} synthesis filters G ₁ and G ₂ are given by 2. The digital signal compression method according to claim 1, wherein a characteristic satisfying the following is given. 3. The digital signal compression method according to claim 1, wherein said cutoff frequency is fs / 4. 4. The digital signal compression method according to claim 1, wherein the operation step of operating the characteristic obtains a predetermined characteristic by changing a characteristic on a frequency axis and a condition for canceling aliasing noise. 5. The digital signal compression method according to claim 1, wherein in the operation step of operating the characteristic, the operation of the characteristic on the frequency axis is prioritized over the condition in which aliasing noise cancels out in a high frequency band. 6. A digital signal compression apparatus for compressing information of a digital signal, comprising: a dividing means for dividing an input signal into at least two bands by an analysis filter; and a spectrum on a frequency axis of the signal divided into the at least two bands. Conversion means for obtaining data; and operating means for operating at least one characteristic of the spectrum data on the frequency axis based on frequency characteristic operation information, wherein the operating means for operating the characteristic comprises: Calculate the size of the above spectral data,
The frequency characteristic operation information is allocated to each of the spectrum data and normalized, and a multiplication coefficient calculation circuit that outputs a coefficient for multiplication is provided.The multiplication result is previously determined by the spectrum data and the coefficient supplied from the multiplication coefficient calculation circuit. An overflow inspection circuit for inspecting whether the value exceeds a predetermined upper limit; and, based on an output of the overflow inspection circuit, when the number of spectrum data on the frequency axis at which an overflow occurs is equal to or more than a predetermined value, A coefficient correction circuit that normalizes the entire coefficient, and by correcting the output of the coefficient correction circuit to a characteristic that maintains symmetry with the cutoff frequency of the analysis filter as a symmetric axis based on the frequency characteristic operation information, In order to maintain the condition that aliasing noise is canceled by band synthesis by the synthesis filter, And aliasing noise reduction circuit for reducing noise returns, the digital signal compression apparatus characterized by having a a multiplier for multiplying the spectral data on the coefficients and the frequency axis that received fixed in the aliasing noise reduction circuit. 7. The dividing means for dividing the input signal into at least two bands comprises a quadrature mirror filter comprising at least a low-pass filter and a high-pass filter, and the operating means for operating the characteristic comprises: The impulse response of the low-pass filter is f ₁ (n), and the Z-transform is F ₁ (n)
Then, Holds, and the impulse response f ₂ of the high-pass filter is given by The filled, further, the synthesis filter _G 1, the impulse response _g 1 of _{G 2, g 2} is [6] 7. The digital signal compression apparatus according to claim 6, wherein a characteristic that satisfies the following condition is provided. 8. The digital signal compression device according to claim 6, wherein said cutoff frequency is fs / 4. 9. A digital signal compression apparatus according to claim 6, wherein said characteristic operation means obtains a predetermined characteristic by changing a characteristic on a frequency axis and a condition for canceling aliasing noise. 10. The digital signal compression apparatus according to claim 6, wherein said characteristic operation means prioritizes operation of characteristics on the frequency axis over conditions for canceling aliasing noise in a high frequency range.