Speech coding using a method of discrete energy separation (DESA)

The use of the Discrete Energy Separation Algorithm (DESA), developed by Kaiser, Quatieri, and Maragos (1993) to code speech signals is investigated. We show that the DESA-demodulated signals can be coded and decoded using a combination of polynomial and sinusoidal approximations. We discuss how the data is subband filtered. Then how each subband is processed to encode and decode the data. A brief discussion is given of the quality of the reconstructed or synthesized speech output.