Text-independent speaker verification

We developed a system to identify users independent of their spoken words. The system used Mel-frequency cepstral coefficients (MFCCs) to obtain a characteristic pattern of the user’s voice which was then analysed by a three layer multi-layer perceptron (MLP) neural network. We obtained accuracies of over 90% using this approach.

The entire project was implemented in MATLAB.

Skand Hurkat
Hardware Engineer