Comparison between the Wavelet Model and the Modulation Components for Speaker Recognition | ||
Iraqi Journal of Information Technology | ||
Article 1, Volume 7, Issue 1, January 2017, Pages 104-113 | ||
Author | ||
Tariq A. Hassan | ||
Abstract | ||
This paper presents a comparison between two multi-resolution models, these are the Wavelet transform and the AM-FM modulation model. The aim is to show which of them will perform well in different circumstances. Both models are tested on text-dependent and text-independent corpora. In our experiment, the two models are acting well with clean data. However, with noisy data the results show that the Wavelet model performance decline more than the AM-FM do. This explain that the AM-FM model although both models adopt the multi-resolution analysis in their process; however, the wavelet model still need some robustness against noise. | ||
Keywords | ||
FM modulation; wavelet transform; Mother function; Multiband demodulation analysis method; Quadrature Mirror Filter | ||
Statistics Article View: 92 PDF Download: 84 |