R. Steele

Audio and video bit-rate reduction

We address the problem of estimating limits of bit-rate reduction for audio and video information. Binary encoding of such signals is performed by a myriad of different techniques which we classify into two categories. The goal of waveform encoders is simply replication of waveforms, whereas parameter encoders attempt to extract and transmit only the basic features necessary for the specific application. Encoders are compared in terms of bits per Nyquist sample (η). Minimizing η requires utilization of source signal statistics and human perception. Accordingly, we discuss audio and video sources and their statistics, the salient properties of hearing and vision, and subjective assessment of quality. A review of the present state of waveform and parameter encoding is presented, summarized by graphs of η versus coder complexity for different qualities of reproduction. Current trends are then extrapolated, new directions anticipated and limiting values of η estimated. In summary, we predict that waveform encoders will be refined to the point where bit rates are nearly optimum, but highly variable. Dramatic reductions will only come about through sophisticated parameter encoding.

A Difference Detection and Correction Scheme for Combating DPCM Transmission Errors

It is possible to detect the presence of transmission errors in the decoded sample sequence of a differential pcm system. The detection procedure is based on the sample-to-sample differences in the received signal. It infers an error whenever an individual difference is substantially greater than the rms difference. On the detection of an error, the sample affected is replaced by the result of a smoothing procedure and subsequent samples are adjusted to compensate for the error propagation inherent in differential pcm. This difference detection and correction (DDC) scheme substantially improves system performance at high error rates.

Sequential Adaptive Predictors for ADPCM Speech Encoders

The sequential gradient estimation predictor is compared in detail to the stochastic approximation predictor, and both are evaluated in an ADPCM codec. A switched predictor having two coefficients is then described for use in a DPCM-AQF codec. This predictor divides the range of the correlation coefficient of the speech signal into zones, and as the correlation coefficient changes zones, the predictor coefficients undergo a substantial modification. By this method the adaptation rate of the predictor is improved, particularly during transitions between unvoiced and voiced sounds.

Hardware for Detection and Partial Correction of PCM Transmission Errors

In pulse code modulation of speech, transmission errors cause changes in the statistical properties of the received samples. Recently published work describes computer simulations of a difference detection and correction (DDC) system which detect errors on the basis of these changes. The system examines the differences between adjacent received samples. If a particular difference exceeds a threshold, which is dependent on the rms value of a block of 64 differences, the quantized sample responsible for this large difference is deemed to be erroneous. A nonlinear filter of the median type is then introduced as a corrector. We now present a comprehensive description of the hardware realization of this system, and present measured results of signal-to-noise ratio (snr) as a function of bit error rate, input power and the error criterion for narrow-band white noise input signals. An improvement of 6 dB in snr is achieved for error rates between 0.2 and 2.0%. The snr results are consistent with subjective impressions of the quality of speech processed by the DDC system.

Sequential gradient estimation predictor for speech signals

A sequential gradient estimation predictor [SGEP] for speech signals is presented. In a given sampling interval, each prediction coefficient in turn is increased and decreased in value by a prescribed amount, while the other coefficients are kept constant. Two predictions are then made and the better enables the coefficient to be modified in the correct direction, but by an amount determined by a number of factors. The superiority of SGEP over the stochastic approximation predictor is illustrated by waveforms and snr (signal to noise ratio) performance curves.

Envelope Dynamic Ratio Quantizer

The envelope dynamic ratio quantizer (envelope DRQ) is an adaptive quantizer for speech signals. By utilizing the envelope of the speech, nonlinear elements, a fixed quantizer and a simple predictor, a closed-loop adaptive quantizer emerges having a high constant SNR over a wide dynamic range. The theory of the quantizer is presented, together with computer simulation results which show an improvement.compared to the one word memory APCM system. Finally, the simplicity of implementing the envelope DRQ is described.