Training "on the fly" to improve the performance of speaker recognition in noisy environments

Al-Noori, AHY; Duncan, PJ; Li, FF

Abstract

Reliability of Speaker Recognition (SR) is crucial for critical applications, especially in adverse acoustic
conditions. Ambient noises and their variations represent a significant challenge for such applications. In this
paper, a new technique is proposed to address the issue of performance degradation in noisy environments.
Based on the estimation of the signal to noise ratio (SNR) and profile of the ambient noise from input signals,
the proposed method re-trains the enrolment model for the claim speaker to generate new noisy models that
adapt to the noise profile. This technique is termed “training on the fly”. Evaluation results show notable
enhancement in performance in terms of the reduction of equal error rates over a range of SNRs and different
types of noise.

Citation

Al-Noori, A., Duncan, P., & Li, F. (2017). Training "on the fly" to improve the performance of speaker recognition in noisy environments. In Proceedings: 2017 AES International Conference on Audio Forensics. Audio Engineering Society

• Start Date: Jun 15, 2017
• End Date: Jun 17, 2017
• Publication Date: Jun 15, 2017
• Deposit Date: Jul 19, 2017
• Publisher: Audio Engineering Society
• Book Title: Proceedings: 2017 AES International Conference on Audio Forensics
• ISBN: 9781942220145
• Publisher URL: http://www.aes.org/e-lib/browse.cfm?elib=18744
• Related Public URLs: http://www.aes.org/; http://www.aes.org/conferences/2017/forensics/
• Additional Information: Event Type : Conference