The number of applications of the Linear Predictive Coding (LPC) method of processing speech sounds in general, has grown enormously. This is largely due to the time domain nature of the analysis and ease of computation. The majority of its applications have been in the linear all-pole modelling of speech although actual speech production is known to produce zeros. Short-time speech spectral matching by pole-zero modelling has been attempted [Atal and Schroeder, 1978, Makhoul, 1975, Yegnanarayana,1981].