Data processing: speech signal processing – linguistics – language – Audio signal bandwidth compression or expansion
Reexamination Certificate
1999-03-12
2004-05-25
Knepper, David D. (Department: 2654)
Data processing: speech signal processing, linguistics, language
Audio signal bandwidth compression or expansion
C704S200100
Reexamination Certificate
active
06741965
ABSTRACT:
TECHNICAL FIELD
This invention relates to coding method and device suitable for expanding the format of coded signals, decoding method and device corresponding thereto, and a recording medium on which coded signals are recorded.
BACKGROUND ART
Conventionally, a signal recording medium such as a magneto-optical disc has been proposed as a medium on which signals like coded acoustic information or audio information (hereinafter referred to as audio signals) can be recorded. There are various methods for high-efficiency coding of the audio signals, which can be exemplified by, for example, so-called transform coding, which is a blocking frequency band division system for blocking audio signals on the time base by a predetermined time unit, then transforming (spectrum transform) the signals on the time base of each block to signals on the frequency base so as to divide the signal into a plurality of frequency bands, and coding the signal of each band, or so-called subband coding (SBC), which is a non-blocking frequency band division system for dividing audio signals on the time base into a plurality of frequency bands without blocking the signals, and then coding the signals. Also, a method for high-efficiency coding using the above-described subband coding and transform coding in combination is considered. In this case, for example, after band division is carried out in the subband coding, the signal of each band is spectrum-transformed to a signal on the frequency base, and this spectrum-transformed signal of each band is coded.
As a filter for band division used in the above-described subband coding, a filter such as a so-called QMF (quadrature mirror filter) is employed. This QMF filter is described in R. E. Crochiere, “Digital coding of speech in subbands,” Bell Syst. Tech. J., Vol.55, No.8, 1976. This QMF filter is adapted for bisecting a band with equal band widths, and is characterized in that so-called aliasing is not generated in synthesizing the divided bands. Also, in Joseph H. Rothweiler, “Polyphase Quadrature filters—A new subband coding technique,” ICASSP 83, BOSTON, a filter division method for equal band widths is described. This polyphase quadrature filter is characterized in that it can divide, at a time, a signal into a plurality of bands of equal band widths.
As the above-described spectrum transform, for example, input audio signals are blocked by a predetermined unit time (frame), and discrete Fourier transform (DFT), discrete cosine transform (DCT), modified discrete cosine transform (MDCT) or the like is carried out for each block, thereby transforming the time base to the frequency base. The above-mentioned MDCT is described in J. P. Princen, A. B. Bradley, “Subband/Transform Coding Using Filter Bank Designs Based on Time Domain Aliasing Cancellation,” Univ. of Surrey Royal Melbourne Inst. of Tech., ICASSP 1987.
In the case where the above-mentioned DFT or DCT is used as a method for spectrum transform of a waveform signal, M units of independent real-number data are obtained by carrying out transform using a time block constituted by M units of sample data. (This block is hereinafter referred to as a transform block.) To reduce connection distortion between transform blocks, normally, M
1
units of sample data of each of adjacent transform blocks are caused to overlap each other. Therefore, in DFT or DCT, M units of real-number data are obtained with respect to (M-M
1
) units of sample data on the average, and these M units of real-number data are subsequently quantized and coded.
On the other hand, in the case where the above-mentioned MDCT is used as a method for spectrum transform, M units of real-number data are obtained from 2M units of sample data which are obtained by causing M units of sample data of each of adjacent transform blocks to overlap each other. That is, in the case where MDCT is used, M units of real-number data are obtained with respect to M units of sample data on the average, and these M units of real-number data are subsequently quantized and coded. In a decoding device, waveform elements obtained by carrying out inverse transform of each block, from the code obtained by using MDCT, are added to each other while being caused to interfere with each other, thereby reconstituting a waveform signal.
Meanwhile, in general, if the transform block for spectrum transform is made long, the frequency resolution is enhanced and energy is concentrated on a specified spectral signal component. Therefore, by carrying out spectrum transform using a long transform block length obtained by causing sample data of adjacent transform blocks to overlap each other by half thereof each, and using MDCT such that the number of obtained spectral signal components is not increased with respect to the number of original sample data on the time base, more efficient coding can be carried out than in the case where DFT or DCT is used. Also, by providing a sufficiently long overlap of the adjacent transform blocks, connection distortion between the transform blocks of the waveform signals can be reduced. However, since a longer transform block for transform requires a greater work area for transform, it becomes an obstacle to miniaturization of reproducing means or the like. Particularly, employment of a long transform block at the time when increase in integration degree of a semiconductor is difficult leads to increase in cost, and therefore needs to be considered carefully.
As described above, by quantizing a signal component divided for each band by using a filter or spectrum transform, a band where quantization noise is generated can be controlled. Therefore, utilizing characteristics of a so-called masking effect, auditorily more efficient coding can be carried out. In addition, by normalizing each sample data using the maximum value of the absolute value of the signal component in each band before carrying out quantization, more efficient coding can be carried out.
As the frequency division width in the case where each signal obtained by carrying out frequency band division of audio signals is to be quantized, a band width in consideration of human auditory characteristics may be preferably used. Specifically, it is preferred to divide audio signals into a plurality of bands (for example, 25 bands) by using a band width referred to as a critical band that generally becomes greater in higher frequency bands. In coding data of each band in this case, coding based on predetermined bit distribution for each band or adaptive bit allocation for each band is carried out. For example, in coding coefficient data obtained by MDCT processing by using the above-mentioned bit allocation, coding with an adaptive number of allocated bits is carried out with respect to MDCT coefficient data of each band obtained by MDCT processing for each transform block. As bit allocation methods, the following two method are known.
For example, in R. Zelinski and P. Noll, “Adaptive Transform Coding of Speech Signals,” IEEE Transactions of Acoustics, Speech, and Signal Processing, vol. ASSP-25, No.4, August 1977, bit allocation is carried out on the basis of the magnitude of the signal of each band. In this method, the quantization noise spectrum becomes flat and the noise energy becomes minimum. However, since the masking effect is not utilized, the actual feeling of noise is not auditorily optimum.
In addition, in M. A. Kransner, “The critical band coder—digital encoding of the perceptual requirements of the auditory system,” MIT, ICASSP 1980, a method for carrying out fixed bit allocation by utilizing auditory masking to obtain a necessary signal-to-noise ratio for each band is described. With this method, however, even in measuring characteristics by sine wave input, the resultant characteristic value is not so satisfactory because bit allocation is fixed.
To solve these problems, there has been proposed a high-efficiency coding method in which all bits that can be used for bit allocation are used in a divided manner for a fixed bit allocation pattern predetermined for each small
Shimoyoshi Osamu
Tsutsui Kyoya
Knepper David D.
Sonneschein Nath & Rosenthal LLP
Sony Corporation
LandOfFree
Differential stereo using two coding techniques does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Differential stereo using two coding techniques, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Differential stereo using two coding techniques will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3244249