Title :
Systolic neural network architecture for second order hidden Markov models
Author :
Zhaozhi, Feng ; Zailu, Huang ; Daowen, Chen ; Faguan, Wan
Author_Institution :
National Lab. of Pattern Recognition, Inst. of Autom., Acad. Sinica, Beijing, China
fDate :
30 Aug-3 Sep 1992
Abstract :
This paper presents a systolic neural network architecture for implementing second order hidden Markov models (SOHMMs). A programmable systolic arrays is proposed. A unified model for recurrent high order multilayer feedforward neural networks and SOHMMs is exploited for the architecture design. Extended Viterbi algorithm for SOHMMs is described. Finally, the implementation based on TMS320C25 chip is also discussed
Keywords :
digital signal processing chips; feedforward neural nets; hidden Markov models; neural chips; parallel architectures; recurrent neural nets; systolic arrays; hidden Markov models; programmable systolic arrays; systolic neural network architecture; Algorithm design and analysis; Feedforward neural networks; Hidden Markov models; Multi-layer neural network; Neural networks; Pattern recognition; Recurrent neural networks; Systolic arrays; Viterbi algorithm; Wide area networks;
Conference_Titel :
Pattern Recognition, 1992. Vol. IV. Conference D: Architectures for Vision and Pattern Recognition, Proceedings., 11th IAPR International Conference on
Conference_Location :
The Hague
Print_ISBN :
0-8186-2925-8
DOI :
10.1109/ICPR.1992.202163
