Pulse or digital communications – Pulse code modulation – Noise or distortion reduction
Patent
1993-10-19
1995-08-22
Chin, Stephen
Pulse or digital communications
Pulse code modulation
Noise or distortion reduction
375285, 375343, H04B 110, H04B 1500
Patent
active
054447410
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to a filtering method and device for reducing the preechoes of a digital audio signal.
2. The Prior Art
The storage, transmission or processing of digitized sound signals, digital audio signals, currently comes up against the problem of the necessary digital transmission rate. Techniques such as coding by frequency transform are then used to allow discrimination of the necessary transmission rate, whilst preserving the quality of the original signal.
However, owing precisely to the enormous variety of sound signals, the aforesaid methods or techniques do not always enable the quality of the original signal to be preserved.
Coding by frequency transform has, hitherto, been used widely for coding high-fidelity sound.
Generally, this type of coding proceeds by slicing the time digital audio signal into blocks of N samples, followed by a weighting by a window of each block, then by a time/frequency transformation delivering a set of coefficients, which are coded and finally transmitted. For a more detailed description of this type of coding, reference may profitably be made to the French Patent Application 89 13649, inventor Yannick MAHIEUX, published under number 2 653 280.
Several frequency transforms may be used, but the Modified Discrete Cosine Transform MDCT, is of particular interest and allows a bigger reduction in the rate, through the use of weighting windows with softened edges, combined with a large overlap of N/2 samples between two successive blocks, as represented in FIG. 1a.
In this type of coding by transform, the coding noise resulting from the quantizing of the coefficients arising from the transform is distributed uniformly throughout the duration of each block. When the sample block contains a nonstationarity such as that caused for example by the sudden plucking of a stringed instrument, see FIG. 1b, the frequency spectrum of the signal is then virtually flat. However, the aforesaid coding processes generally carry out a spectral shaping of the noise, see aforesaid French patent application. Thus, for a block containing a sudden transition, the noise spectrum is virtually flat and of constant level throughout the duration of the block.
Accordingly, as regards the part of the digital audio signal preceding the plucking, the noise spectrum is much greater than that of the signal and, as represented in FIG. 1c, in the time domain, the decline constituting the preecho phenomenon may be very substantial.
When, on the other hand, the coding system uses the MDCT transform, the two blocks preceding the transition are affected with preechoes, owing to the overlap of N/2 samples between successive windows.
Among the methods proposed for reducing or eliminating preechoes, one of them advocates the use of scale factors applied to the sample block before processing by frequency transform, so as to reduce the difference in level between the parts before and after the transition.
On decoding, by applying inverse scale factors, the level of the noise is reduced in the areas of low energy and the preecho phenomenon diminishes accordingly. For a more complete description reference may be made for example to the article by Sugiyama: Adaptive Transform Coding with an adaptive Block Size, Proceedings of ICASSP 90, Albuquerque, pp 1093-1096.
Such a method, when an MDCT transform is used, cannot however be employed since, owing to the scale factors, it is not possible to undertake, coding excepted, perfect reconstruction of the digital audio signal.
A second method, from the aforesaid methods, can consist in the use of blocks of variable length, the size of the transform, for a sample block containing a sudden transition, being reduced to N/8 points, which makes it possible thus to limit the duration of the decline, see for example the document ISO-IEC-WG8-MPEG 89/205.
Although this second method may be employed when an MDCT transform is used, the sample block weighting windows must be modified. Furthermore, it is necessary to detect th
REFERENCES:
patent: 5311549 (1994-05-01), Mahieux
Bocure Tesfaldet
Chin Stephen
France Telecom
LandOfFree
Filtering method and device for reducing digital audio signal pr does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Filtering method and device for reducing digital audio signal pr, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Filtering method and device for reducing digital audio signal pr will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2147602