Multitaper chirp group delay Hilbert envelope coefficients for robust speaker verification

摘要

In this paper, we propose two new feature extraction methods for robust automatic speaker verification under noisy conditions. The first method, called Multi-taper Gammatone Hilbert Envelope Coefficients (MGHECs), employs multi-taper magnitude spectra that offer considerable advantages for spectrum estimates. The second method, called Multi-taper Chirp Group Delay Zeros-Phase Hilbert Envelope Coefficients (MCGDZPHECs) based on multi-tapers phase spectral. The chirp group delay technique is used to estimate the vocal tract from the chirp Fourier transform phase. The performance evaluation of the proposed methods and their extended variants are carried out on NIST 2008 corpus under noisy conditions, using various noise SNR levels which are extracted from NOISEX-92. Experimental results show that the proposed methods provide better representation of speech spectrum. Moreover, we obtained a significant improvement in performance under noisy conditions when compared to conventional Mean Hilbert Envelope Coefficients (MHECs) feature extraction.