Detection of Disguised Voice Using Probabilistic Neural Network

Abstract

Since voice disguise is the process of concealing one’s identity, it is being widely used for illegal purposes. This has caused a negative impact on the audio forensics, so it is important to identify whether the voice is disguised or not. It is more difficult to identify whether the voice is disguised or not, if the voice is disguised using electronic scrambling devices or audio editing software tools. We know that voice disguise is the modification of frequency spectrum of speech signals, so we will be using mel-frequency cepstrum coefficients (MFCC’s). In this paper, we will be extracting MFCC statistical moments including mean and correlation coefficients as acoustic features and then we will be using an algorithm based on these features and will be using probabilistic neural network (PNN) as classifier to distinguish whether the voice is disguised or not.
DOI: 10.17762/ijritcc2321-8169.150806