Performance Analysis of Feature Extraction Methods in Indoor Sound Classification

Abstract

In this paper, by using a novel database of home enviroment warning sounds, the classification and recognition performances of these sounds are compared over feature extraction algorithms. Following the sample reduction of the feature vectors by Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), k-Nearest Neighbour (k-NN) algorithm is employed for classification. Besides, a modified version of the algorithm for MF coefficients is proposed and we observe that the classification performance is better than MFCC and LPC even at low SNR values.