Header

UZH-Logo

Maintenance Infos

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 (2018). A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition. Frontiers in Neuroscience, 12:160.

Abstract

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.

Abstract

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.

Statistics

Citations

Dimensions.ai Metrics
5 citations in Web of Science®
5 citations in Scopus®
Google Scholar™

Altmetrics

Downloads

19 downloads since deposited on 08 Mar 2019
7 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Journal Article, refereed, original work
Communities & Collections:07 Faculty of Science > Institute of Neuroinformatics
Dewey Decimal Classification:570 Life sciences; biology
Scopus Subject Areas:Life Sciences > General Neuroscience
Language:English
Date:2018
Deposited On:08 Mar 2019 16:35
Last Modified:01 Jul 2021 14:20
Publisher:Frontiers Research Foundation
ISSN:1662-453X
OA Status:Gold
Free access at:PubMed ID. An embargo period may apply.
Publisher DOI:https://doi.org/10.3389/fnins.2018.00160
PubMed ID:29643760

Download

Gold Open Access

Download PDF  'A comparison of low-complexity real-time feature extraction for neuromorphic speech recognition'.
Preview
Content: Published Version
Filetype: PDF
Size: 2MB
View at publisher
Licence: Creative Commons: Attribution 4.0 International (CC BY 4.0)