Speech Recognition Systems and Its Automatic Synthesis in Hardware

In this paper a methodology to perform optimizations in the hardware implementation of a speech recognizer is presented; this methodology explores the design space from a very high level representation of the algorithm. We expose how to accelerate the process of speech recognition by exploiting the inherently concurrent decoding stage. Finally, we show the cost in terms of hardware resources required to synthesize the speech recognition algorithms.