Low-power neuromorphic speech recognition engine with coarse-grain sparsity

Shihui Yin; Deepak Kadetotad; Bonan Yan; Chang Song; Yiran Chen; Chaitali Chakrabarti; Jae-sun Seo

doi:10.1109/ASPDAC.2017.7858305

Low-power neuromorphic speech recognition engine with coarse-grain sparsity

Shihui Yin, Deepak Kadetotad, Bonan Yan, Chang Song, Yiran Chen, Chaitali Chakrabarti, Jae-sun Seo

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

6 Scopus citations

Abstract

In recent years, we have seen a surge of interest in neuromorphic computing and its hardware design for cognitive applications. In this work, we present new neuromorphic architecture, circuit, and device co-designs that enable spike-based classification for speech recognition task. The proposed neuromorphic speech recognition engine supports a sparsely connected deep spiking network with coarse granularity, leading to large memory reduction with minimal Index information Simulation results show that the proposed deep spiking neural network accelerator achieves phoneme error rate (PER) of 20.5% for TIMIT database, and consume 2.57mW in 40nm CMOS for real-time performance. To alleviate the memory bottleneck, the usage of non-volatile memory is also evaluated and discussed.

Original language	English (US)
Title of host publication	2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	111-114
Number of pages	4
ISBN (Electronic)	9781509015580
DOIs	https://doi.org/10.1109/ASPDAC.2017.7858305
State	Published - Feb 16 2017
Event	22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017 - Chiba, Japan Duration: Jan 16 2017 → Jan 19 2017

Publication series

Name	Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

Other

Other	22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017
Country/Territory	Japan
City	Chiba
Period	1/16/17 → 1/19/17

ASJC Scopus subject areas

Electrical and Electronic Engineering
Computer Science Applications
Computer Graphics and Computer-Aided Design

Access to Document

10.1109/ASPDAC.2017.7858305

Cite this

Yin, S., Kadetotad, D., Yan, B., Song, C., Chen, Y., Chakrabarti, C., & Seo, J. (2017). Low-power neuromorphic speech recognition engine with coarse-grain sparsity. In 2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017 (pp. 111-114). Article 7858305 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ASPDAC.2017.7858305

Low-power neuromorphic speech recognition engine with coarse-grain sparsity. / Yin, Shihui; Kadetotad, Deepak; Yan, Bonan et al.
2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 111-114 7858305 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Yin, S, Kadetotad, D, Yan, B, Song, C, Chen, Y, Chakrabarti, C & Seo, J 2017, Low-power neuromorphic speech recognition engine with coarse-grain sparsity. in 2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017., 7858305, Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, Institute of Electrical and Electronics Engineers Inc., pp. 111-114, 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017, Chiba, Japan, 1/16/17. https://doi.org/10.1109/ASPDAC.2017.7858305

Yin S, Kadetotad D, Yan B, Song C, Chen Y, Chakrabarti C et al. Low-power neuromorphic speech recognition engine with coarse-grain sparsity. In 2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 111-114. 7858305. (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). doi: 10.1109/ASPDAC.2017.7858305

@inproceedings{60cdbbe22fab4cd6af0db96bfd9bb8ef,

title = "Low-power neuromorphic speech recognition engine with coarse-grain sparsity",

abstract = "In recent years, we have seen a surge of interest in neuromorphic computing and its hardware design for cognitive applications. In this work, we present new neuromorphic architecture, circuit, and device co-designs that enable spike-based classification for speech recognition task. The proposed neuromorphic speech recognition engine supports a sparsely connected deep spiking network with coarse granularity, leading to large memory reduction with minimal Index information Simulation results show that the proposed deep spiking neural network accelerator achieves phoneme error rate (PER) of 20.5% for TIMIT database, and consume 2.57mW in 40nm CMOS for real-time performance. To alleviate the memory bottleneck, the usage of non-volatile memory is also evaluated and discussed.",

author = "Shihui Yin and Deepak Kadetotad and Bonan Yan and Chang Song and Yiran Chen and Chaitali Chakrabarti and Jae-sun Seo",

note = "Funding Information: This work was partially supported by National Science Foundation grant 1535669 and Intel Corporation.; 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017 ; Conference date: 16-01-2017 Through 19-01-2017",

year = "2017",

month = feb,

day = "16",

doi = "10.1109/ASPDAC.2017.7858305",

language = "English (US)",

series = "Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "111--114",

booktitle = "2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017",

}

TY - GEN

T1 - Low-power neuromorphic speech recognition engine with coarse-grain sparsity

AU - Yin, Shihui

AU - Kadetotad, Deepak

AU - Yan, Bonan

AU - Song, Chang

AU - Chen, Yiran

AU - Chakrabarti, Chaitali

AU - Seo, Jae-sun

N1 - Funding Information: This work was partially supported by National Science Foundation grant 1535669 and Intel Corporation.

PY - 2017/2/16

Y1 - 2017/2/16

N2 - In recent years, we have seen a surge of interest in neuromorphic computing and its hardware design for cognitive applications. In this work, we present new neuromorphic architecture, circuit, and device co-designs that enable spike-based classification for speech recognition task. The proposed neuromorphic speech recognition engine supports a sparsely connected deep spiking network with coarse granularity, leading to large memory reduction with minimal Index information Simulation results show that the proposed deep spiking neural network accelerator achieves phoneme error rate (PER) of 20.5% for TIMIT database, and consume 2.57mW in 40nm CMOS for real-time performance. To alleviate the memory bottleneck, the usage of non-volatile memory is also evaluated and discussed.

AB - In recent years, we have seen a surge of interest in neuromorphic computing and its hardware design for cognitive applications. In this work, we present new neuromorphic architecture, circuit, and device co-designs that enable spike-based classification for speech recognition task. The proposed neuromorphic speech recognition engine supports a sparsely connected deep spiking network with coarse granularity, leading to large memory reduction with minimal Index information Simulation results show that the proposed deep spiking neural network accelerator achieves phoneme error rate (PER) of 20.5% for TIMIT database, and consume 2.57mW in 40nm CMOS for real-time performance. To alleviate the memory bottleneck, the usage of non-volatile memory is also evaluated and discussed.

UR - http://www.scopus.com/inward/record.url?scp=85015345993&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85015345993&partnerID=8YFLogxK

U2 - 10.1109/ASPDAC.2017.7858305

DO - 10.1109/ASPDAC.2017.7858305

M3 - Conference contribution

AN - SCOPUS:85015345993

T3 - Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

SP - 111

EP - 114

BT - 2017 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 22nd Asia and South Pacific Design Automation Conference, ASP-DAC 2017

Y2 - 16 January 2017 through 19 January 2017

ER -

Low-power neuromorphic speech recognition engine with coarse-grain sparsity

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this