A Configurable Logic Based Architecture for Real-Time ContinuousSpeech Recognition Using Hidden Markov Models

  • Authors:
  • Panagiotis Stogiannos;Apostolos Dollas;Vassilis Digalakis

  • Affiliations:
  • Technical University of Crete, 73100 Chania, Crete, Greece;Technical University of Crete, 73100 Chania, Crete, Greece;Technical University of Crete, 73100 Chania, Crete, Greece

  • Venue:
  • Journal of VLSI Signal Processing Systems - Special issue on VLSI on custom computing technology
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

An architecture is presented for real-time continuousspeech recognition based on a modified hidden Markov model. Thealgorithm is adapted to the needs of continuous speech recognition byefficient encoding of the state space, and logarithmic encoding ofthe weights so that products can be computed as sums. The paperpresents the algorithm and its application related modifications, themapping of the algorithm to a special purpose architecture, and thedetailed design of this architecture using configurable logic.Emphasis is given on how the attributes of the algorithm areexploited in a configurable logic based design. A concrete designexample is presented with a coprocessor engine having one large FPGA,64 Mbytes of synchronous DRAM (SDRAM), a small FPGA as a SDRAMcontroller, and 2 Mbytes SRAM. This engine operating at 66 MHzperforms roughly nine times as fast as a high end personal computerrunning a fully optimized version of the same algorithm.