Method of encoding a signal

Details Method of encoding a signal Method of encoding a signal

2000-02-16

2003-04-08

Decady, Albert (Department: 2133)

Error detection/correction and fault detection/recovery

Pulse or data error handling

Digital data error correction

Reexamination Certificate

active

06546515

ABSTRACT:

BACKGROUND OF THE INVENTION
This invention relates to a method of encoding a signal, particularly an analog, time-varying signal for speech and/or image transmission. The encoding is used, for example, in transmitting equipment of a communications system. The communications system is, for example, a wireline or wireless system or a hybrid system. Conventional wireline systems are, for example, telephone networks, cable television networks, or the like. Conventional wireless systems are, for example, radio communications systems, radio relay systems, mobile radio systems, GSM networks, CDMA networks, UMTS networks, DECT networks, or satellite-based networks, such as “Iridium”.
In present-day digital communications systems, the analog signal provided by the data source, e.g., speech=sound, audio, or image=video, is subjected at the transmitting end to a data compression in order to reduce the redundancy contained in the signal and thus minimize the bandwidth needed for transmission. This signal processing method is referred to as “source coding”.
Source coding for speech is generally based on a model of the human vocal tract, and only those current parameters of the model which describe the current speech are transmitted to the receiver. There, the speech is synthesized by means of the parameters.
Similar signal processing methods are used in image transmission, where a set of pixels is subjected to a transform, and the transform coefficients as well as motion vectors and control signals are transmitted as parameters.
During data transmission, parametric quantities (signal values, filter coefficients, amplification factors, etc.) are determined, which are then quantized and encoded into sequences of binary digits. The number of quantization levels of the individual parameters is frequently a power of two.
During the transmission of the quantized parameters over real transmission channels, errors arise. Errors in particular bits, so-called sensitive bits, of the sequence generally result in a more severe degradation of the received signal quality than errors in less sensitive bits. Therefore, the bits of the source encoder are grouped in sensitivity classes, for example, for which different degrees of error protection are used in the form of error-correcting codes. This method is generally referred to as channel coding with unequal error protection (UEP).
By the source coding methods, the parametric quantities to be transmitted are, as a rule, quantized in such a way that the signal-to noise ratio (SNR) at the transmitting end is maximized for the available bit rate. The number of possible quantization levels is frequently a power of two 2
M
, so that the M bits used for source coding can be fully utilized.
SUMMARY OF THE INVENTION
Furthermore, source coding and channel coding are optimized largely separately.
It is an object of the invention to provide a novel method of encoding a signal which comprises source and/or channel coding.
The invention relates to a method whereby the nonintegral excess share of M−Id(Q) bits during the (source) coding of a quantized parameter can be systematically used to increase error robustness, i.e., during the quantization of an amplitude-continuous parameter with a quantizer whose number of levels is not a power of two, the nonuse of some binary indices (binary patterns) is to be used at the transmitting end to provide systematic error protection, with table decoding being preferably used at the receiving end in conjunction with a parameter estimator. A code redundancy, e.g., an algebraically constructed code redundancy, is no longer decoded by conventional channel decoding using some kind of a hard or soft decision, but indirectly by means of parameter estimation. Thus, error robustness can be further increased by adding a code redundancy at the transmitting end, such as an error correcting-code. During the parameter estimation at the receiving end, this code redundancy is incorporated into the decoding.
It frequently turns out that a scalar or vectorial quantization into 2
M−1
<Q<2
M
levels is sufficient to guarantee the required basic quality, e.g., the SNR, of the compression process. In prior-art methods, the remaining effective word length of M−Id(Q) bits (Id=logarithm to the base 2) per parameter causes an improvement in the basic quality at the transmitting end, which may not be necessary. If only a minimum quality is required, this is suboptimal in terms of the quality received over a noisy channel, since the proportionate data rate, corresponding to M−Id(Q) bits per parameter, should be turned to better account; it should be used, for example, to increase error robustness.
In prior-art methods in which a redundancy-increasing quantization is achieved by leaving some binary indices unused, this is not done with a view to systematically increasing error robustness.
At the transmitting end, prior-art methods do not take into account that instead of parameter decoding, an estimation of the parameter value can be performed at the receiving end, whereby error robustness is further improved, see Patent Application DE 197 16 147.2.
In prior-art methods, complex channel decoding is used at the receiving end to decode the redundancy of an error-correcting code.
Some advantages of the invention are as follows.
The association between quantized parameter values and bit codes to be transmitted is chosen optimally to meet a suitable criterion regarding parameter characteristics and channel characteristics which takes into account the characteristics and sensitivities of the source signal.
Any number of Q<2
M
quantization levels can be encoded with M bits in such a way that the redundancy of M−Id(Q) bits can be used to provide optimum error protection by minimizing the mean error between the input and output of the communications system. In particular, Q need not be power of two.
By varying the number of quantization levels, the share of useful information and error-protecting redundancy in the gross bit stream can be adjusted in fine steps. If required, an additional redundancy of a code can be added.
Because of the signal-matched coding (source and/or channel coding), all bits used to encode a parameter can then have approximately the same low sensitivity to transmission errors.
The error protection is parameter-, not bit-oriented. As a result, the residual redundancy still present after source coding can be used to increase error robustness and thus achieve a higher transmission quality.
The novel method supports the receive-side parameter estimation already on the transmit side in optimum fashion, so that significant gains can be achieved over conventional decoding.


REFERENCES:
patent: 5091945 (1992-02-01), Kleijn
patent: 197 16 147 (1998-10-01), None
patent: 0 529 909 (1993-03-01), None
patent: 0 643 493 (1995-03-01), None
Fingscheidt et al., Speech Decoding with Error Concealment using Residual source Redundancy, IEEE, p. 91-92, 1997.*
Phamdo et al., Source-dependent Channel Coding of CELP Speech over Land Mobile Radio Channels, IEEE, p. 1041-1045, 1995.*
Heinen S.; Adratm M.; Steil O; Vary P; Xu W.: “A 6.1 to 13.3-kb/s variable rate CELP codec (VR-CELP) for AMR speech coding” 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99, Bd. 1, Nr. 1, Mar. 15-19, 1999, Seiten 9-12, XP002121916.

Affiliated with

Also associated with

Rating

Method of encoding a signal does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Method of encoding a signal, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of encoding a signal will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3081627