Environment Sound Recognition for Digital Audio Forensics Using ZC, MFCC, MPEG-7 and LPC Features

摘要

In this paper, we perform several experiments focusing on the problems of environment recognition from audio particularly for forensic application. We investigated the effect of temporal Zero Crossing feature, Mel Frequency Cepstral Coefficients feature, some selected MPEG-7 audio low level descriptors feature and Linear Predictive Coding feature on environment sound recognition. The performance is evaluated against varying number of training sounds and samples per each training file. Experimental results show that higher recognition accuracy is achieved by increasing the number of training files and by decreasing the number of samples per training file.