Data processing: speech signal processing – linguistics – language – Speech signal processing – For storage or transmission
Reexamination Certificate
1999-05-04
2002-03-05
Dorvil, Richemond (Department: 2741)
Data processing: speech signal processing, linguistics, language
Speech signal processing
For storage or transmission
C704S201000, C704S224000, C704S500000
Reexamination Certificate
active
06353807
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method and an apparatus for transforming coded information signals. The invention also relates to a program providing medium on which a program for transforming coded information signals is recorded.
2. Description of the Related Art
High-efficiency coding methods are known in which the amount of data for audio or sound signals is compressed with very little loss in the acoustic quality. Various high-efficiency coding methods for coding audio or sound signals are available and include, for example, the non-block frequency band division technique, i.e., the subband coding (SBC) technique, and the block frequency-band division technique, i.e., the transform coding technique. In the subband coding technique, an audio signal in the time domain is divided into a plurality of frequency bands and coded rather than forming the audio signal into blocks. In the transform coding technique, a signal in the time domain is transformed (spectrum-transformed) into a signal in the frequency domain, and is divided into a plurality of frequency bands. The signal component in each band is then coded.
As a filter used for frequency division, a quadrature mirror filter (QMF) may be used, which is discussed in the technical document “R. E. Crochiere,
Digital coding of speech in subbands,
Bell Syst. Tech. J., Vol. 55, No. 8, 1976”.
In the above-described QMF filter, aliasing components generated by signals decimated to a half rate after performing frequency division are canceled by aliasing components generated when the signals in the respective bands are synthesized. Because of this characteristic, the loss incurred by coding can be almost completely eliminated if the signal components in the respective bands are coded with a sufficiently high precision.
In the technical document “Joseph H. Rothweiler,
Polyphase Quadrature filters—A new Subband coding technique,
ICASSP 83, BOSTON, 1983”, a polyphase quadrature filter (PQF) filter used in the equal-bandwidth filter division technique is described. In this PQF filter, aliasing components generated by the signal components between the adjacent bands, which are decimated to a rate in accordance with the bandwidth after performing frequency division, are canceled by aliasing components generated by the signal components between the adjacent bands when the signal components in the respective bands are synthesized. Because of this characteristic, the loss incurred by coding can be almost completely eliminated if the signal components in the respective bands are coded with a sufficiently high precision.
As the aforementioned spectrum transform, the following type of spectrum transform, for example, is known. An input audio signal is formed into blocks with a predetermined unit time (frame), and discrete Fourier transform (DFT), discrete cosine transform (DCT), modified discrete cosine transform (MDCT), etc. may be performed on the signal component in each block, thereby transforming the time domain signal into signal components in the frequency domain. The MDCT is discussed in, for example, the technical document “J. P. Princen, A. B. Bradley,
Subband/Transform Coding Using Filter Bank Designs Based on Time Domain Aliasing Cancellation,
ICASSP 1987, Univ. of Surrey Royal Melbourne Inst. of Tech.”.
According to the above-described DFT or DCT used as a method for transforming a waveform signal into a spectrum, the waveform signal is transformed by a time block having M number of samples so as to obtain M independent items of real-number data. In order to reduce the connection distortion between the time blocks, samples are generally overlapped by M
1
number of samples between adjacent blocks. Accordingly, in the DFT or DCT, M items of real-number data are quantized and coded in relation to (M-M
1
) number of samples.
On the other hand, according to the above-described MDCT used as a method for transforming a waveform signal into a spectrum, M independent items of real-number data are obtained from 2M samples having an overlapping portion between adjacent time blocks of M number of samples. Accordingly, in the MDCT, M items of real-number data are quantized and coded in relation to M samples. For example, in a decoder, the codes obtained by using the MDCT are inverse-transformed in the respective blocks so as to produce waveform elements. The waveform elements are then added while interfering with each other, thereby reconstructing the waveform signal.
Generally, by increasing the length of the time block used for transforming, the frequency resolution of the spectra is increased to concentrate energy in a specific spectral component. Thus, by using the MDCT in which transforming is conducted with a longer block having an overlapping portion by an amount of a half block between adjacent blocks, and in which the obtained number of spectral signal components is not more than the original number of time samples, coding can be performed with higher efficiency than in the aforementioned DFT or DCT.
Additionally, a sufficient length of the overlapping portions is provided between adjacent blocks, thereby reducing the interblock distortion of a waveform signal. However, a larger work area for transforming is required with an increased length of the transform block, thereby hampering the miniaturization of, for example, reproduction means. This causes an increase in cost, particularly when it is difficult to increase the integration level of a semiconductor.
According to the above description, signal components divided into the respective bands by using a filter or spectrum transform are quantized, which makes it possible to control the bands in which quantizing noise is generated. By further utilizing characteristics, such as the masking effect, acoustically higher-efficiency coding can be performed.
The masking effect is an effect in which louder sounds acoustically mask softer sounds. By utilizing this effect, the generated quantizing noise can be acoustically masked by the original signal sound. Thus, the sound quality of the compressed signal is almost the same as that of the original signal. For effectively utilizing the masking effect, however, it is necessary to control the generation of the quantizing noise in the time domain or in the frequency domain. For example, if quantizing noise is generated for a few microseconds or greater during a small magnitude of signal immediately before the attack portion in which the magnitude of signal sharply increases, it is no longer masked by the signal sound. This further brings about the loss of sound quality to such a degree as to be uncomfortable from an auditory point of view, which is referred to as “pre-echo”. To overcome this drawback, the block length used in transforming a waveform signal into spectral signal components is changed in accordance with the characteristics of the signal component in the corresponding block. Before performing quantization, the signal component in each band is normalized by the maximum of the absolute value of the signal component, thereby making it possible to perform higher-efficiency coding.
To determine the frequency division width used for quantizing the individual frequency components, a band division technique may be employed by considering human auditory characteristics. For example, in a bandwidth that increases toward the higher range, which is generally referred to as the “critical band”, the band division technique for dividing an audio signal into a plurality of, for example, 25 bands may be employed.
Data in each band is coded by performing predetermined bit allocation or adaptive bit allocation. For example, in coding coefficient data obtained by the aforementioned MDCT according to the above-described bit allocation, the MDCT coefficient data in each band obtained by performing the aforementioned MDCT on each block is coded with the adaptively allocated number of bits.
As the bit allocation methods, the following two methods are known.
In a first method, bit allocation is c
Shimoyoshi Osamu
Tsutsui Kyoya
Dorvil Richemond
Sonnenschein Nath & Rosenthal
Sony Corporation
LandOfFree
Information coding method and apparatus, code transform... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Information coding method and apparatus, code transform..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Information coding method and apparatus, code transform... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2843334