An Empirical Study of a Wavelet-Neural Network Based Approach for Forensic Speaker Recognition with Cross Channel Data
Claude Turner, Dwight H. Richards, Joseph, Anthony · ASME Press eBooks · 2010
This paper presents a neural network and wavelet based approach for forensic speaker recognition with cross channel data. We discuss our algorithm and provide a performance comparison between the proposed system and a cepstral feature extraction and Gaussian mixture model (GMM) based approach. Our results show that the proposed system performs as well as or better than the cepstral-GMM based model for semi-text-dependent speech and the same set of six speakers from the TIMIT and CTIMIT corpora.