Joint Sound Source Separation and Speaker Recognition

2016/04/29 by Jeroen Zegers, Zegers, Jeroen, Hugo Van hamme +1
Computer Science · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Sound (cs.SD) #cs.LG #cs.SD

paper · pdf · doi:10.48550/arxiv.1604.08852

Submitted to INTERSPEECH2016. 4 pages, 1 extra page for references

arxiv created 2016/04/29 · arxiv updated 2016/05/02

Abstract

Non-negative Matrix Factorization (NMF) has already been applied to learn speaker characterizations from single or non-simultaneous speech for speaker recognition applications. It is also known for its good performance in (blind) source separation for simultaneous speech. This paper explains how NMF can be used to jointly solve the two problems in a multichannel speaker recognizer for simultaneous speech. It is shown how state-of-the-art multichannel NMF for blind source separation can be easily extended to incorporate speaker recognition. Experiments on the CHiME corpus show that this method outperforms the sequential approach of first applying source separation, followed by speaker recognition that uses state-of-the-art i-vector techniques.

Related

• Monaural Audio Speaker Separation with Source Contrastive Estimation
• Target Speech Extraction Based on Blind Source Separation and X-vector-based Speaker Selection Trained with Data Augmentation
• End-to-end Non-Negative Autoencoders for Sound Source Separation
• Multichannel blind speech source separation with a disjoint constraint source model
• A Neural Network Alternative to Non-Negative Audio Models