Performance Comparison of Affine Projection Algorithm and Normalized Kernel Affine Projection Algorithm for Speech Enhancement

The main objective of the Speech Enhancement system is to improve the intelligibility and the quality of degraded speech signal. This paper emphasize on enhancement of noisy speech by using Affine Projection Algorithm (APA) and Kernel Affine Projection Algorithm (KAPA). Noise is present everywhere in the environment, So Kernel adaptive filters are used to enhance noisy speech signal and shows the good results in increasing the signal to Noise Ratio (SNR) and minimizing the mean square error (MSE). The computer simulations are performed using NOIZEUS speech corpus for different SNR values using Affine projection (APA), Kernel Affine projection (KAPA), Kernel normalized least  mean square (KNLMS), Kernel affine projection with coherent criteria (KAPCC), and Extended kernel recursive least square (EXKRLS) their performance is compared.