Enhancing Text-Independent Speaker Verification through Advanced Deep Representation Feature Analysis with Neural Networks

Speaker Verification (SV) identifies individuals II. RELATED WORK by evaluating their distinctive vocal qualities. Text-independent speaker authentication is an important task in speech recognition and security applications. This proposes an extended fine-grained analysis of deep representation features for text-independent speaker verification using neural networks. This paper uses a Combination of CNN and RNN Neural Network Models for verification. Two Machine Learning Models Random Forest Classifier and AdaBoost Classifier was implemented on speaker verification system, but the neural network model outperforms both the models. The model achieves the accuracy of 96.4%. The approach has numerous practical uses across different security and access control systems, as well as in speech recognition technologies.