"Speech Recognition on Vector Architectures"
From cellphones and PDAs to huge automated call centers, speech recognition is becoming more and more ubiquitous. As demand for automatic speech recognition (ASR) applications increases, so too does the need to run ASR algorithms on a variety of unconventional computer architectures. One such architecture uses a vector processor, which has many benefits in terms of performance, power consumption, price, etc.
This talk presents and evaluates ASR algorithms ported to run efficiently on vector architectures. A vector simulation library was developed, and is used to evaluate design trade-offs for both the algorithms and the hardware.