Discrete type dynamic synapses neural network: Speech recognition application
Hassan H. Namarvar, Jim‐Shih Liaw, Theodore W. Berger · The Journal of the Acoustical Society of America · 2001
A discrete type of the Dynamic Synapses Neural Network (DSNN) has been developed and applied to speech recognition. In order to speed up the training time of the network, a new discrete time implementation of the original DSNN [J.-S. Liaw and T. W. Berger, 1996] has been introduced based on the impulse invariant transformation. The new architecture of the network was trained with the Genetic Algorithms [H. H. Namarvar et al., 2001] and tested against the continues-type DSNN. The overall speed of the new algorithm with discrete-time difference equation set is about 13 times faster than the same algorithm with the continuous differential equation set. This significant reduction of processing time not only decreases the training time but also makes the system better suited for real-time speech recognition tasks. [Work supported by DARPA.]