Speaker recognition using neural networks

Abstract

This project demonstrates how to build a speaker recognition system using deep learning techniques. The system uses MFCC to extract features from audio data and capture spectral and time domain information of speech. After comparing traditional classification methods and neural network classification methods, then chooses a recurrent neural network (RNNs) to process of sequence data using. The project was trained and evaluated on the famous audio dataset VoxCelebl to train and evaluate various speaker recognition models using python. The system achieved a test accuracy of 93%. Tlus result demonstrates tliat the system is able to effectively distinguish between different speakers.