Memory-Efficient Modeling and Search Techniques for Hardware ASR Decoders

Michael Price, Anantha Chandrakasan, James Glass


This paper gives an overview of acoustic modeling and search techniques for low-power embedded ASR decoders. Our design decisions prioritize memory bandwidth, which is the main driver in system power consumption. We evaluate three acoustic modeling approaches — Gaussian mixture model (GMM), subspace GMM (SGMM) and deep neural network (DNN) — and identify tradeoffs between memory bandwidth and recognition accuracy. We also present an HMM search scheme with WFST compression and caching, predictive beam width control, and a word lattice. Our results apply to embedded system implementations using microcontrollers, DSPs, FPGAs, or ASICs.


DOI: 10.21437/Interspeech.2016-287

Cite as

Price, M., Chandrakasan, A., Glass, J. (2016) Memory-Efficient Modeling and Search Techniques for Hardware ASR Decoders. Proc. Interspeech 2016, 1893-1897.

Bibtex
@inproceedings{Price+2016,
author={Michael Price and Anantha Chandrakasan and James Glass},
title={Memory-Efficient Modeling and Search Techniques for Hardware ASR Decoders},
year=2016,
booktitle={Interspeech 2016},
doi={10.21437/Interspeech.2016-287},
url={http://dx.doi.org/10.21437/Interspeech.2016-287},
pages={1893--1897}
}