Speech coding by the Hadamard-Haar transform

This paper reports on the combined application of the Hadamard and Haar transforms as a method of speech coding. The results of an extensive investigation of the properties of 64-point Hadamard-Haar transformed speech are presented, with detailed information being provided about the probability-density functions of the Haar coefficients, about the average power-density spectrum in the Haar domain, and about the autocorrelation function of speech reconstructed from a limited number of Haar coefficients. A preliminary series of listening tests have been performed and they confirm the conclusions drawn from the statistical properties of the transformed speech. The preliminary listening tests indicate that intelligible, though somewhat noisy, speech can be obtained at bit rates of approximately 3.5kbit/s.