Comparison of LPC algorithms for real-time speech processing

摘要

Several methods for solving Yule-Walker equations, including the autocorrelation, covariance, lattice, and covariance-lattice algorithms, are investigated. Computational efficiency, memory requirements, stability, optimality, spectral resolution and fixed-point suitability of the algorithms are studied and compared. Such an analysis is required by real-time fixed-point implementations of linear predictive coding (LPC) of speech signals.