On the use of TDNN-extracted features information in talker identification
ICASSP '91 Proceedings of the Acoustics, Speech, and Signal Processing, 1991. ICASSP-91., 1991 International Conference
A tutorial on text-independent speaker verification
EURASIP Journal on Applied Signal Processing
IEEE Transactions on Audio, Speech, and Language Processing
Hi-index | 0.00 |
A modified GMM with an embedded TDNN is proposed to speaker recognition. The model integrates the merits of GMM and TDNN. TDNN is used to digest the time information of the feature sequences, and through the transformation of the feature vectors the model makes the hypothesis of variable independence which maximum likelihood needed more reasonable. In the process of training, GMM and TDNN are trained as a whole and the parameters of GMM and TDNN are updated alternately. Experiments show that the proposed model improves accuracy rate against baseline GMM at all SNR with a maximum to 22%.