Error detection/correction and fault detection/recovery – Pulse or data error handling – Digital data error correction
Reexamination Certificate
1999-11-22
2003-07-29
Dildine, R. Stephen (Department: 2133)
Error detection/correction and fault detection/recovery
Pulse or data error handling
Digital data error correction
C708S404000
Reexamination Certificate
active
06601206
ABSTRACT:
CROSS-REFERENCE TO RELATED APPLICATION
This application claims priority of European Patent Application No. 98309935.9, which was filed on Dec. 4, 1998.
FIELD OF THE INVENTION
This invention relates to methods of and apparatus for error concealment or error correction of signals such as speech, image and video signals, and particularly for such signals as transmitted over wireless and ATM channels.
BACKGROUND OF THE INVENTION
In Marvasti and Nafie, “Sampling Theorem: A Unified Outlook on Information Theory, Block and Convolutional Codes”, IEICE Trans. Fundamentals, vol E76-A, no. 9, September 1993 (hereinafter referred to as “ST”) and Marvasti, Hasan and Echhart, “Speech Recovery with Bursty Losses”, Proc of ISCAS '98, Monterey, Calif., USA, May 31-Jun. 3, 1998 (hereinafter referred to as “BL”) there is disclosed a technique for error concealment or correction of an analogue signal using the discrete Fourier transform (DFT). The signal is sampled and formed into blocks and the DFT is taken. Then the DFT is modified so that M consecutive components are equal to zero (bearing in mind the cyclic nature of the coefficients in a DFT, so that for a block length of N the Nth coefficient and the first are consecutive). For error correction, these components are inserted into the DFT spectrum, so that the block size is greater than the original number of samples, whereas for error concealment, M components of the DFT are set to zero, so that some information in the signal is lost. Then the inverse Fourier transform is taken and the resulting new signal is transmitted, after the usual frame processing. As pointed out in ST, the positions of the zero-value components can be, and preferably are, selected so that the inverse DFT gives a signal with real values, assuming that the original signal has real values. In ST it is shown that it is possible to correct up to M erasures per block, and some methods for doing so are described. If the block length of the transmitted signal is N and the number of coefficients set to zero is M and N=K+M, possible methods involve inverting a K×K matrix, and interpolating the missing values using Lagrange interpolation. Both of these methods become computationally intensive for large N, and BL discloses a recursive method which is much more tractable. Unfortunately, as is pointed out in BL, rounding errors accumulate when the block size becomes large, especially when the erasures that are being corrected appear in consecutive samples, and it has been found that the recovery methods become unstable for block sizes greater than N=64.
To comply with telecommunications standards such as the G.729 or GSM standards, samples need to be transmitted in frames of a certain size, which is 80 samples in the case of G.729 and 160 samples in the case of GSM. Since transmission errors causing erasures cause whole frames to be dropped, the block length must be a multiple of the frame length. In other words, block lengths must be greatly in excess of N=64. Furthermore, loss of a frame means loss of 80 or 160 consecutive samples, so the stability of the error recovery method is under considerable pressure.
It is an object of the present invention to provide a technique for error concealment or error correction which has improved stability for large block sizes.
SUMMARY OF THE INVENTION
The invention is based on the discovery of a modified form of discrete Fourier transform which is formally similar to the normal DFT in that the same recovery, methods can be used as in ST and BL, but which does not suffer from the instability to the same degree. Furthermore, when actually performing the transform and the inverse transform, the conventional FFT method can be used, enabling much existing hardware and software to be employed in the modified technique.
More specifically, the method of the present invention comprises a method (and apparatus) for introducing increased redundancy into a discrete signal for the purpose of error concealment or correction of bursty and/or frame losses, the method comprising the steps of forming the discrete Fourier transform of a block of samples of said signal; modifying the Fourier transform to form a block of coefficients of block length N in which a predetermined set of M Fourier coefficients are equal to zero; and forming the inverse discrete Fourier transform of the block of Fourier coefficients to form the increased-redundancy signal. More particularly, the predetermined set of Fourier coefficients consists of the coefficients numbered p(k) where k takes on M successive integer values, p(k) being defined as p(k)=pk mod N where p is a positive integer, greater than unity, relatively prime to N.
The present invention also comprises a method (and apparatus) for providing error concealment or correction of a sampled analog signal in which a signal is recovered from a received signal in which some samples have not been correctly received by using redundancy introduced into the transmitted signal, the redundancy consisting of a predetermined set of M Fourier components in each block of the discrete Fourier transform of the transmitted signal of block length N being equal to zero. Again, the predetermined set of Fourier coefficients consists of the coefficients numbered p(k) where k takes on M successive integer values, p(k) being defined as p(k)=pk mod N where p is a positive integer, greater than unity, relatively prime to N.
REFERENCES:
patent: 3605019 (1971-09-01), Cutter et al.
patent: 6404817 (2002-06-01), Saha et al.
patent: 1006665 (2000-06-01), None
patent: 2 753 592 (1996-09-01), None
patent: 57067347 (1982-04-01), None
Farokh Marvasti et al, “Sampling Theorem: A Unified Outlook On Information Theory, Block and Convolutional Codes”, IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences, vol. 76A, No. 9, Sep. 1, 1993, pp. 1383-1391, Tokyo.
European Search Report, 98309935,9-2209-, May 14, 1999.
Agere Systems Inc.
Dildine R. Stephen
LandOfFree
Error concealment or correction of speech, image and video... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Error concealment or correction of speech, image and video..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Error concealment or correction of speech, image and video... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3015813