Analysis of some redundancy removal bandwidth compression techniques

Three redundancy removal bandwidth compression algorithms --the floating-aperture predictor, the zero-order interpolator, and the fan interpolator--are analyzed. Theoretical expressions are found for the mean and mean-square times between output samples of these devices when the input signal is a Markov process. These expressions are evaluated for the case in which the input is a first-order Gaussian Markov process, and the resulting output sampling rates and transmission bandwidths are compared to those required by a PCM system using uniform sampling and optimum linear filter interpolation. It is shown that, given sufficient a priori knowledge of the signal process, there is little to be gained by using these redundancy removal techniques in place of the PCM system. However, if the signal statistics are unknown, the use of these algorithms instead of PCM may provide a considerable bandwidth reduction.