AN IMPROVED SPEAKER VERIFICATION SYSTEM FOR ROBOTIC APPLICATIONS

Abstract

Text Dependent Human Voice Recognition (TDHVR) systems are used to verify the identity of individuals based on their speech signals. This abstract presents a TDHVR system that utilizes statistical computation, formant estimation, and wavelet energy analysis to achieve accurate verification. The system is evaluated using fifty preloaded voice signals from six individuals, and the proposed algorithm achieves an accuracy rate of approximately 90%, surpassing the performance of Linear Predictive Coding (LPC), which achieves only 66.66% accuracy.

Through extensive simulation tests conducted on various speech signals from different speakers, it is observed that the proposed algorithm significantly improves the accuracy of the TDHVR system compared to LPC. The integration of statistical computation, formant estimation, and wavelet energy analysis enhances the system's ability to accurately verify the identity of individuals based on their speech signals.

This work contributes to the field of voice recognition by presenting a novel approach that outperforms existing techniques, such as LPC. The achieved accuracy rate of approximately 90% demonstrates the effectiveness and potential of the proposed algorithm in practical applications requiring reliable identity verification through speech signals.