LPC excitation based on zinc function decomposition

The quality of low-bit-rate speech using linear prediction is largely dependent on the model used for the excitation signal. The authors propose a linear predictive coding (LPC) excitation model composed of a set of orthogonal functions called zinc functions, which are well-suited for modeling the LPC residual signal. The zinc basis functions are used in a low-bit-rate, multiphase LPC speech-coding system. Results show that, given a fixed segmental signal-to-noise ratio, the zinc multipulse LPC (ZMPLPC) system is more efficient than a conventional multipulse LPC (MPLPC) system, with similar computational complexity; the saving in transmitted information is achieved at a minimal increase in the number of computations. Subjective listening tests also indicate a preference for the ZMPLPC system