On a delta modulation based real time autocorrelator

Theoretical analysis indicates that the quantization error of a linear delta modulator (LDM) output can be divided into an equivalent PCM quantization error and a high frequency noise irrelevant to the input. Under certain treatment the autocorrelation of a DM output can approximate that of the input signal´s derivative. Simulation and experimental results prove that a delta modulation based autocorrelator (DMAC) is an efficient aid to extracting speech features.