Speaker Verification: Attention Based Deep Learning Architecture
Naman Gadag, Rai Khatawate, Satish Chikkamath, S. R. Nirmala, Suneeta V. Budihal · 2024
Speaker verification verifies the claimed identity of a person through speech signal. Speaker verification methods are classified into two types text dependent and the other is text independent. Speaker verification systems play a crucial role in security systems. Deep learning has been widely applied in speaker verification, Deep learning methods have shown significant improvements in various aspects of speaker verification. Some of them are CNN, RNN, LSTM. This paper introduces an attention-based model for speaker verification. The proposed model leverages attention mechanisms to focus on the speaker speech. Experimental results demonstrate efficacy of attention-based model showcasing the improved verification performance compared to normal speaker verification.