The term “autocorrelation method” encompasses the inverse filter, maximum likelihood, and PARCOR method. This method has a significant computational advantage over Atal's covariance method since the analysis does not require polynominal root testing programs and associated modifications, and is thus much easier to implement. Informal listening tests indicate that for low bitrate coding, the resulting synthetic speech is very similar for both methods with fixed frame analysis. Directly from the autocorrelation equation solution, the filter parameters, k parameters (also referred to as reflection coefficients and PARCOR parameters) are obtained, along with the synthesizer gain parameter. Low bit-rate coding based upon the k parameters has been obtained by performing the following steps: (1) prefilter sharply at the half-sampling frequency; (2) prewhiten and Hamm before analysis; (3) transform the parameter values to [ −0.7, 0.7]; (4) quantize with rounding; and (5) interpolate synthesis parameters pitch synchronously. These operations are all very fast and easily implemented. Good quality synthesis over a 3200-Hz input bandwidth is obtainable with [ki BITS, i = 0,1, ⋯,7,] = [6,6,3, ⋯,3]. Examples coded at 2100 and 1400 bits/sec (50-and 33-Hz frame rates, respectively) are presented.