A micropower analog circuit implementation of hidden Markov model state decoding

We describe the implementation of a hidden Markov model state decoding system, a component for a wordspotting speech recognition system. The key specification for this state decoder design is microwatt power dissipation: this requirement led to a continuous-time, analog circuit implementation. We describe the tradeoffs inherent in the choice of an analog design and explain the mapping of the discrete-time state decoding algorithm into the continuous domain. We characterize the operation of a ten-word (81-state) state decoder test chip.

[1]  T. Delbruck.  'Bump' circuits for computing similarity and dissimilarity of analog voltages , 1991, IJCNN-91-Seattle International Joint Conference on Neural Networks.

[2]  Hervé Bourlard,et al.  Connectionist Speech Recognition: A Hybrid Approach , 1993 .

[3]  H. Oguey,et al.  CMOS Current Reference without Resistance , 1996, ESSCIRC '96: Proceedings of the 22nd European Solid-State Circuits Conference.

[4]  Marwan A. Jabri,et al.  A low-power network for on-line diagnosis of heart patients , 1995, IEEE Micro.

[5]  Richard F. Lyon,et al.  Improved implementation of the silicon cochlea , 1992 .

[6]  Richard Lippmann,et al.  Wordspotter training using figure-of-merit back propagation , 1994, Proceedings of ICASSP '94. IEEE International Conference on Acoustics, Speech and Signal Processing.

[7]  T. W. Matthews,et al.  An integrated analog CMOS Viterbi detector for digital magnetic recording , 1993 .

[8]  André van Schaik,et al.  Improved Silicon Cochlea using Compatible Lateral Bipolar Transistors , 1995, NIPS.