A 23-μW Keyword Spotting IC With Ring-Oscillator-Based Time-Domain Feature Extraction

Kwantae Kim; Chang Gao; Rui Graca; Ilya Kiselev; Hoi Jun Yoo; Tobi Delbruck; Shih Chii Liu

doi:10.1109/JSSC.2022.3195610

A 23-μW Keyword Spotting IC With Ring-Oscillator-Based Time-Domain Feature Extraction

Kwantae Kim, Chang Gao, Rui Graca, Ilya Kiselev, Hoi Jun Yoo, Tobi Delbruck, Shih Chii Liu^*

^*Corresponding author for this work

Electronics

Research output: Contribution to journal › Article › Scientific › peer-review

12 Citations (Scopus)

99 Downloads (Pure)

Abstract

This article presents the first keyword spotting (KWS) IC that uses a ring-oscillator-based time-domain processing technique for its analog feature extractor (FEx). Its extensive usage of time-encoding schemes allows the analog audio signal to be processed in a fully time-domain manner except for the voltage-to-time conversion stage of the analog front end. Benefiting from fundamental building blocks based on digital logic gates, it offers better technology scalability compared to conventional voltage-domain designs. Fabricated in a 65-nm CMOS process, the prototyped KWS IC occupies 2.03 mm 2 and dissipates 23- $\mu \text{W}$ power consumption, including analog FEx and digital neural network classifier. The 16-channel time-domain FEx achieves a 54.89-dB dynamic range for 16-ms frame shift size while consuming 9.3 $\mu \text{W}$. The measurement result verifies that the proposed IC performs a 12-class KWS task on the Google Speech Command dataset (GSCD) with >86% accuracy and 12.4-ms latency.

Original language	English
Pages (from-to)	3298-3311
Number of pages	14
Journal	IEEE Journal of Solid-State Circuits
Volume	57
Issue number	11
DOIs	https://doi.org/10.1109/JSSC.2022.3195610
Publication status	Published - 2022

Bibliographical note

Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Keywords

Analog
bandpass filter (BPF)
classifier
feature extractor (FEx)
Google Speech Command dataset (GSCD)
keyword spotting (KWS)
rectifier
recurrent neural network (RNN)
ring oscillator
time domain

Access to Document

10.1109/JSSC.2022.3195610

A_23-W_Keyword_Spotting_IC_With_Ring-Oscillator-Based_Time-Domain_Feature_ExtractionFinal published version, 6.08 MB

Cite this

@article{a7f8ad3acf554350a71b6b605ca908bb,

title = "A 23-μW Keyword Spotting IC With Ring-Oscillator-Based Time-Domain Feature Extraction",

abstract = "This article presents the first keyword spotting (KWS) IC that uses a ring-oscillator-based time-domain processing technique for its analog feature extractor (FEx). Its extensive usage of time-encoding schemes allows the analog audio signal to be processed in a fully time-domain manner except for the voltage-to-time conversion stage of the analog front end. Benefiting from fundamental building blocks based on digital logic gates, it offers better technology scalability compared to conventional voltage-domain designs. Fabricated in a 65-nm CMOS process, the prototyped KWS IC occupies 2.03 mm 2 and dissipates 23- $\mu \text{W}$ power consumption, including analog FEx and digital neural network classifier. The 16-channel time-domain FEx achieves a 54.89-dB dynamic range for 16-ms frame shift size while consuming 9.3 $\mu \text{W}$. The measurement result verifies that the proposed IC performs a 12-class KWS task on the Google Speech Command dataset (GSCD) with >86% accuracy and 12.4-ms latency. ",

keywords = "Analog, bandpass filter (BPF), classifier, feature extractor (FEx), Google Speech Command dataset (GSCD), keyword spotting (KWS), rectifier, recurrent neural network (RNN), ring oscillator, time domain",

author = "Kwantae Kim and Chang Gao and Rui Graca and Ilya Kiselev and Yoo, {Hoi Jun} and Tobi Delbruck and Liu, {Shih Chii}",

note = "Green Open Access added to TU Delft Institutional Repository {\textquoteleft}You share, we take care!{\textquoteright} – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.",

year = "2022",

doi = "10.1109/JSSC.2022.3195610",

language = "English",

volume = "57",

pages = "3298--3311",

journal = "IEEE Journal of Solid-State Circuits",

issn = "0018-9200",

publisher = "IEEE",

number = "11",

}

TY - JOUR

T1 - A 23-μW Keyword Spotting IC With Ring-Oscillator-Based Time-Domain Feature Extraction

AU - Kim, Kwantae

AU - Gao, Chang

AU - Graca, Rui

AU - Kiselev, Ilya

AU - Yoo, Hoi Jun

AU - Delbruck, Tobi

AU - Liu, Shih Chii

N1 - Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

PY - 2022

Y1 - 2022

N2 - This article presents the first keyword spotting (KWS) IC that uses a ring-oscillator-based time-domain processing technique for its analog feature extractor (FEx). Its extensive usage of time-encoding schemes allows the analog audio signal to be processed in a fully time-domain manner except for the voltage-to-time conversion stage of the analog front end. Benefiting from fundamental building blocks based on digital logic gates, it offers better technology scalability compared to conventional voltage-domain designs. Fabricated in a 65-nm CMOS process, the prototyped KWS IC occupies 2.03 mm 2 and dissipates 23- $\mu \text{W}$ power consumption, including analog FEx and digital neural network classifier. The 16-channel time-domain FEx achieves a 54.89-dB dynamic range for 16-ms frame shift size while consuming 9.3 $\mu \text{W}$. The measurement result verifies that the proposed IC performs a 12-class KWS task on the Google Speech Command dataset (GSCD) with >86% accuracy and 12.4-ms latency.

AB - This article presents the first keyword spotting (KWS) IC that uses a ring-oscillator-based time-domain processing technique for its analog feature extractor (FEx). Its extensive usage of time-encoding schemes allows the analog audio signal to be processed in a fully time-domain manner except for the voltage-to-time conversion stage of the analog front end. Benefiting from fundamental building blocks based on digital logic gates, it offers better technology scalability compared to conventional voltage-domain designs. Fabricated in a 65-nm CMOS process, the prototyped KWS IC occupies 2.03 mm 2 and dissipates 23- $\mu \text{W}$ power consumption, including analog FEx and digital neural network classifier. The 16-channel time-domain FEx achieves a 54.89-dB dynamic range for 16-ms frame shift size while consuming 9.3 $\mu \text{W}$. The measurement result verifies that the proposed IC performs a 12-class KWS task on the Google Speech Command dataset (GSCD) with >86% accuracy and 12.4-ms latency.

KW - Analog

KW - bandpass filter (BPF)

KW - classifier

KW - feature extractor (FEx)

KW - Google Speech Command dataset (GSCD)

KW - keyword spotting (KWS)

KW - rectifier

KW - recurrent neural network (RNN)

KW - ring oscillator

KW - time domain

UR - http://www.scopus.com/inward/record.url?scp=85136861304&partnerID=8YFLogxK

U2 - 10.1109/JSSC.2022.3195610

DO - 10.1109/JSSC.2022.3195610

M3 - Article

AN - SCOPUS:85136861304

SN - 0018-9200

VL - 57

SP - 3298

EP - 3311

JO - IEEE Journal of Solid-State Circuits

JF - IEEE Journal of Solid-State Circuits

IS - 11

ER -

A 23-μW Keyword Spotting IC With Ring-Oscillator-Based Time-Domain Feature Extraction

Abstract

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this