Title :
Real time neural networks. III. Alternative neural networks for speech applications
Author :
Tatman, Günhan ; Jannarone, R.
Author_Institution :
Dept. of Electr. & Comput. Eng., South Carolina Univ., Columbia, SC, USA
Abstract :
For pt.II, see ibid., p.584-90 (1991). Waibel Sawai and Shikano (1989), have successfully used traditional neural networks (TNNs), in conjunction with a certain time-delay framework, to recognize phonemes in spoken Japanese syllables. In this paper, the authors first describe Waibel et. al.´s time delay neural network model. They then make several modifications to that model and indicate how procedures can be developed to automatically identify key variables and their cross products, within the structured conjunctoid neural net framework rather than the nonparametric TNN framework,
Keywords :
neural nets; speech recognition; conjunctoid neural net; speech recognition; spoken Japanese syllables; time delay neural network model; traditional neural networks; Data mining; Delay effects; Equations; Fires; Logic devices; Logistics; Neural networks; Neurons; Speech; Training data;
Conference_Titel :
System Theory, 1991. Proceedings., Twenty-Third Southeastern Symposium on
Conference_Location :
Columbia, SC
Print_ISBN :
0-8186-2190-7
DOI :
10.1109/SSST.1991.138636
