Title :
TDNN based speaker identification
Author :
Berger, Gary L. ; Gowdy, John N.
Author_Institution :
Dept. of Electr. & Comput. Eng., Clemson Univ., SC, USA
Abstract :
A spatio-temporal network the time delay neural network (TDNN) is considered. Because the TDNN is time-shift invariant, input patterns do not need to be aligned. Also, the TDNN uses temporal information in its mapping process. Experimental results show that a TDNN-based speaker identification outperforms a weighted-distance-measure-based speaker identification system that does not use temporal information
Keywords :
delay circuits; learning (artificial intelligence); neural nets; speaker recognition; temporal reasoning; mapping; spatio-temporal network; speaker identification; temporal information; time delay neural network; Backpropagation; Biological system modeling; Brain modeling; Data mining; Feature extraction; Linear predictive coding; Neural networks; Pattern classification; Speech; Statistics;
Conference_Titel :
System Theory, 1993. Proceedings SSST '93., Twenty-Fifth Southeastern Symposium on
Conference_Location :
Tuscaloosa, AL
Print_ISBN :
0-8186-3560-6
DOI :
10.1109/SSST.1993.522810
