A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition
This paper presents a real-time, low-complexity neuromorphic speech recognition system using a spiking silicon cochlea, a feature extraction module and a population encoding method based Neural Engineering Framework (NEF)/Extreme Learning Machine (ELM) classifier IC. Several feature extraction metho...
Saved in:
Main Authors: | , , , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2018
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/85120 http://hdl.handle.net/10220/45134 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-85120 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-851202020-11-01T04:44:22Z A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition Acharya, Jyotibdha Patil, Aakash Li, Xiaoya Chen, Yi Liu, Shih-Chii Basu, Arindam Interdisciplinary Graduate School (IGS) School of Electrical and Electronic Engineering Silicon Cochlea Neural Engineering Framework This paper presents a real-time, low-complexity neuromorphic speech recognition system using a spiking silicon cochlea, a feature extraction module and a population encoding method based Neural Engineering Framework (NEF)/Extreme Learning Machine (ELM) classifier IC. Several feature extraction methods with varying memory and computational complexity are presented along with their corresponding classification accuracies. On the N-TIDIGITS18 dataset, we show that a fixed bin size based feature extraction method that votes across both time and spike count features can achieve an accuracy of 95% in software similar to previously report methods that use fixed number of bins per sample while using ~3× less energy and ~25× less memory for feature extraction (~1.5× less overall). Hardware measurements for the same topology show a slightly reduced accuracy of 94% that can be attributed to the extra correlations in hardware random weights. The hardware accuracy can be increased by further increasing the number of hidden nodes in ELM at the cost of memory and energy. Published version 2018-07-19T06:15:48Z 2019-12-06T15:57:29Z 2018-07-19T06:15:48Z 2019-12-06T15:57:29Z 2018 Journal Article Acharya, J., Patil, A., Li, X., Chen, Y., Liu, S.-C., & Basu, A. (2018). A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition. Frontiers in Neuroscience, 12, 160-. 1662-4548 https://hdl.handle.net/10356/85120 http://hdl.handle.net/10220/45134 10.3389/fnins.2018.00160 en Frontiers in Neuroscience © 2018 Acharya, Patil, Li, Chen, Liu and Basu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. 15 p. application/pdf |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Silicon Cochlea Neural Engineering Framework |
spellingShingle |
Silicon Cochlea Neural Engineering Framework Acharya, Jyotibdha Patil, Aakash Li, Xiaoya Chen, Yi Liu, Shih-Chii Basu, Arindam A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
description |
This paper presents a real-time, low-complexity neuromorphic speech recognition system using a spiking silicon cochlea, a feature extraction module and a population encoding method based Neural Engineering Framework (NEF)/Extreme Learning Machine (ELM) classifier IC. Several feature extraction methods with varying memory and computational complexity are presented along with their corresponding classification accuracies. On the N-TIDIGITS18 dataset, we show that a fixed bin size based feature extraction method that votes across both time and spike count features can achieve an accuracy of 95% in software similar to previously report methods that use fixed number of bins per sample while using ~3× less energy and ~25× less memory for feature extraction (~1.5× less overall). Hardware measurements for the same topology show a slightly reduced accuracy of 94% that can be attributed to the extra correlations in hardware random weights. The hardware accuracy can be increased by further increasing the number of hidden nodes in ELM at the cost of memory and energy. |
author2 |
Interdisciplinary Graduate School (IGS) |
author_facet |
Interdisciplinary Graduate School (IGS) Acharya, Jyotibdha Patil, Aakash Li, Xiaoya Chen, Yi Liu, Shih-Chii Basu, Arindam |
format |
Article |
author |
Acharya, Jyotibdha Patil, Aakash Li, Xiaoya Chen, Yi Liu, Shih-Chii Basu, Arindam |
author_sort |
Acharya, Jyotibdha |
title |
A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
title_short |
A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
title_full |
A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
title_fullStr |
A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
title_full_unstemmed |
A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
title_sort |
comparison of low-complexity real-time feature extraction for neuromorphic speech recognition |
publishDate |
2018 |
url |
https://hdl.handle.net/10356/85120 http://hdl.handle.net/10220/45134 |
_version_ |
1683493243615969280 |