Machine Learning based Voice Authentication and Identification

Voice authentication and speech recognition is a technology that enables the identification of certain users or individuals by analyzing their unique voice patterns and characteristics. The voice authentication is a kind of biometric authentication that basically analyses and studies several voice characteristics such as tone, speech patterns, pitch, and speaking rhythm. The technique compares the input voice recorded with the predefined database to authenticate the user and identify them. There are numerous research and studies on voice authentication using various deep learning and CNN techniques. The proposed methodology consists of collecting audio samples and training several various models to distinguish between real and fake users. In this study, results demonstrate the efficiency and effectiveness of various approaches such as FBanksNet, spectrogram-based networks by evaluating accuracy of each technique. FBanksNet slightly performs better due to less training and testing loss.