Window optimization in linear prediction analysis
Wai C. Chu · IEEE Transactions on Speech and Audio Processing · 2003
The autocorrelation method of linear prediction (LP) analysis relies on a window for data extraction. We propose an approach to optimize the window which is based on gradient-descent. It is shown that the optimized window can enhance the performance of LP-based speech coding algorithms; in most instances, improvement in performance comes at no additional computational cost, since it merely requires a window replacement.