Two dimensional prediction and interpolation for data rate compression of LPC parameters

The purpose of this research was to investigate the use of two-dimensional predictive coding of area functions for data rate reduction in LPC-based speech coding systems. Predictive coding in the area function domain is a technique for further data rate reduction on the LPC parameters. Two objective speech distortion measures were used for the optimization of predictor parameters and for objectively assessing the speech quality. The superiority of the LAR domain for coding was verified. The use of two-dimensional predictive coding was shown to be superior to the earlier one-dimensional systems under the conditions tested. Two-dimensional interpolation using a minimum mean square error (MMSE) criterion for interpolation coefficients was shown to be less effective than standard linear interpolation under the conditions tested. However, a one-dimensional nonlinear interpolation using MMSE criterion coefficients was shown comparable to standard linear interpolation.