An integrated speech processing method utilizing general Kalman filter

Speech processing has two major applications namely, Speech enhancement which deals with extracting of clean speech from a noisy one and Speaker identification, a process of providing authentication basing on human speech. Since, speech is a non-stationary signal there are many adaptive algorithms developed to deal with the above applications. One such highly popular algorithm having many forms is Kalman filter. The purpose of present work is to study the use of General Kalman Filter (GKF) a form derived from Kalman filter along with Estimate Maximization (EM) algorithm. The first step of process involves understanding GKF and EM algorithm, second is to use them for Speech enhancement and third step is to use them in pre-processing step of Speaker identification, which shows its significance. Simulation results show the combination of GKF and EM provides a better Speech enhancement and acts as an efficient pre-processing step of Speaker identification over conventional method.