Text-Independent Voice Authentication System Using MFCC Features

Voice authentication is the process of identifying the user based on their voice. There have been many successful attempts with remarkable accuracies. Since, more and more information is getting available on internet, and the secure authentication has become ever more necessary. In this voice authentication system (VAS), English pangrams are used to capture every alphabet utterance of the user. MFCC features are extracted from the collected audio files and a convolutional neural network (CNN) model is trained. While testing, the system generates two random words and concatenates it with the English pangrams to increase the security of the system. This gives the attacker no opportunity to pre-record the sentence and played back against the system to gain the access.