Speaker Identification System To Enhance Real-Time Security Using Convolutional Neural Network

The speaker authentication is important for many domains such as safe keying systems, voice aids, or even criminal investigations and others. The shortcomings of conventional approaches for speaker extraction derive from subjectively determined parameters and empirical models employed. This paper re...

Full description

Saved in:

Bibliographic Details
Published in:	2023 International Conference on Electrical, Communication and Computer Engineering (ICECCE) pp. 1 - 6
Main Authors:	T, Girther, A, Yashwanth Krishnan, G, Ceralaathan, S, Sai Sailesh, VS, Rakshith Raj, V, Rajalakshmi
Format:	Conference Proceeding
Language:	English
Published:	IEEE 30-12-2023
Subjects:	Acoustics Convolutional Neural Network Convolutional neural networks Deep learning Deep Neural Network Real-time systems Security Speaker recognition Training
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	The speaker authentication is important for many domains such as safe keying systems, voice aids, or even criminal investigations and others. The shortcomings of conventional approaches for speaker extraction derive from subjectively determined parameters and empirical models employed. This paper responds to these constraints by constructing the proposed deep learning-speaker recognition system using CNN model. A dataset of speech recordings in 16 KHz sample rate PCM format will be used for training that system. To make the system more robust, we will mimic real-life scenarios by adding noise backgrounds. This study intends to design a deep learning model able to identify various voice features and discriminate voices from noise for speaker recognition. By tuning the CNN model, it will be possible to identify specific traits in each speaker's voice and differentiate among individuals using their unique voices. The emphasis would be placed on building a practical lightweight model suitable for mobile based implementations in real-time speaker ID scenarios.
DOI:	10.1109/ICECCE61019.2023.10442699