Systolic neural network architecture for second order hidden Markov models

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.<
>