Design and evaluation of an ultra low-power human-quality speech recognition system
Visualitza/Obre
Cita com:
hdl:2117/334564
Tipus de documentArticle
Data publicació2020-11
Condicions d'accésAccés obert
Llevat que s'hi indiqui el contrari, els
continguts d'aquesta obra estan subjectes a la llicència de Creative Commons
:
Reconeixement 4.0 Internacional
ProjecteCoCoUnit - CoCoUnit: An Energy-Efficient Processing Unit for Cognitive Computing (EC-H2020-833057)
Abstract
Automatic Speech Recognition (ASR) has experienced a dramatic evolution since pioneer development of Bell Lab’s single-digit recognizer more than 50 years ago. Current ASR systems have taken advantage of the tremendous improvements in AI during the past decade by incorporating Deep Neural Networks into the system and pushing their accuracy to levels comparable to that of humans. This article describes and characterizes a representative ASR system with state-of-the-art accuracy and proposes a hardware platform capable of decoding speech in real-time with a power dissipation close to 1 Watt. The software is based on the so-called hybrid approach with a vocabulary of 200K words and RNN-based language model re-scoring, whereas the hardware consists of a commercially available low-power processor along with two accelerators used for the most compute-intensive tasks. The article shows that high performance can be obtained with very low power, enabling the deployment of these systems in extremely power-constrained environments such as mobile and IoT devices.
CitacióPinto, D.; Arnau, J.; González, A. Design and evaluation of an ultra low-power human-quality speech recognition system. "ACM transactions on architecture and code optimization", Novembre 2020, vol. 17, núm. 4, article 41, p. 1-19.
ISSN1544-3566
Versió de l'editorhttps://dl.acm.org/doi/10.1145/3425604
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
3425604.pdf | 1,170Mb | Visualitza/Obre |