Acoustic scene identification for audio authentication

Abstract

This paper discusses acoustic scene identification which is a part of audio authentication. The work focuses on extracting different acoustical parameters embedded in the recording that can aid in identifying the claimed environment in a legal investigation. Four different feature sets have been used. Out of these, two feature sets have been extracted from the original speech recording and the other two are based on the reverberation component embedded in the recording. Also, the methodology has been tested on two different classifiers. The methodology has been tested on synthetically created speech recordings in seven different environments. The paper gives a comparative study of classification accuracy obtained with different feature sets and different classifiers.