Low-power architecture for epileptic seizure detection based on reduced complexity DWT

Mrigank Sharad, Sumeet Kumar Gupta, Shriram Raghunathan, Pedro P. Irazoqui, Kaushik Roy

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

In this article, we present a low-power, user-programmable architecture for discrete wavelet transform (DWT) based epileptic seizure detection algorithm. A simplified, low-pass filter (LPF)-only-DWT technique is employed in which energy contents of different frequency bands are obtained by subtracting quasiaveraged, consecutive LPF outputs. Training phase is used to identify the range of critical DWT coefficients that are in turn used to set patient-specific system level parameters for minimizing power consumption. The proposed optimizations allow the design to work at significantly lower power in the normal operation mode. The systemhas been tested on neural data obtained from kainate-treated rats. The design was implemented in TSMC-65nm technology and consumes less than 550-nW power at 250-mV supply.

Original languageEnglish (US)
Article number10
JournalACM Journal on Emerging Technologies in Computing Systems
Volume8
Issue number2
DOIs
StatePublished - Jun 1 2012

Fingerprint

Discrete wavelet transforms
Low pass filters
Frequency bands
Rats
Electric power utilization

All Science Journal Classification (ASJC) codes

  • Software
  • Hardware and Architecture
  • Electrical and Electronic Engineering

Cite this

Sharad, Mrigank ; Gupta, Sumeet Kumar ; Raghunathan, Shriram ; Irazoqui, Pedro P. ; Roy, Kaushik. / Low-power architecture for epileptic seizure detection based on reduced complexity DWT. In: ACM Journal on Emerging Technologies in Computing Systems. 2012 ; Vol. 8, No. 2.
@article{d8fcd364981e4ee7952c11cabd26d60e,
title = "Low-power architecture for epileptic seizure detection based on reduced complexity DWT",
abstract = "In this article, we present a low-power, user-programmable architecture for discrete wavelet transform (DWT) based epileptic seizure detection algorithm. A simplified, low-pass filter (LPF)-only-DWT technique is employed in which energy contents of different frequency bands are obtained by subtracting quasiaveraged, consecutive LPF outputs. Training phase is used to identify the range of critical DWT coefficients that are in turn used to set patient-specific system level parameters for minimizing power consumption. The proposed optimizations allow the design to work at significantly lower power in the normal operation mode. The systemhas been tested on neural data obtained from kainate-treated rats. The design was implemented in TSMC-65nm technology and consumes less than 550-nW power at 250-mV supply.",
author = "Mrigank Sharad and Gupta, {Sumeet Kumar} and Shriram Raghunathan and Irazoqui, {Pedro P.} and Kaushik Roy",
year = "2012",
month = "6",
day = "1",
doi = "10.1145/2180878.2180882",
language = "English (US)",
volume = "8",
journal = "ACM Journal on Emerging Technologies in Computing Systems",
issn = "1550-4832",
publisher = "Association for Computing Machinery (ACM)",
number = "2",

}

Low-power architecture for epileptic seizure detection based on reduced complexity DWT. / Sharad, Mrigank; Gupta, Sumeet Kumar; Raghunathan, Shriram; Irazoqui, Pedro P.; Roy, Kaushik.

In: ACM Journal on Emerging Technologies in Computing Systems, Vol. 8, No. 2, 10, 01.06.2012.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Low-power architecture for epileptic seizure detection based on reduced complexity DWT

AU - Sharad, Mrigank

AU - Gupta, Sumeet Kumar

AU - Raghunathan, Shriram

AU - Irazoqui, Pedro P.

AU - Roy, Kaushik

PY - 2012/6/1

Y1 - 2012/6/1

N2 - In this article, we present a low-power, user-programmable architecture for discrete wavelet transform (DWT) based epileptic seizure detection algorithm. A simplified, low-pass filter (LPF)-only-DWT technique is employed in which energy contents of different frequency bands are obtained by subtracting quasiaveraged, consecutive LPF outputs. Training phase is used to identify the range of critical DWT coefficients that are in turn used to set patient-specific system level parameters for minimizing power consumption. The proposed optimizations allow the design to work at significantly lower power in the normal operation mode. The systemhas been tested on neural data obtained from kainate-treated rats. The design was implemented in TSMC-65nm technology and consumes less than 550-nW power at 250-mV supply.

AB - In this article, we present a low-power, user-programmable architecture for discrete wavelet transform (DWT) based epileptic seizure detection algorithm. A simplified, low-pass filter (LPF)-only-DWT technique is employed in which energy contents of different frequency bands are obtained by subtracting quasiaveraged, consecutive LPF outputs. Training phase is used to identify the range of critical DWT coefficients that are in turn used to set patient-specific system level parameters for minimizing power consumption. The proposed optimizations allow the design to work at significantly lower power in the normal operation mode. The systemhas been tested on neural data obtained from kainate-treated rats. The design was implemented in TSMC-65nm technology and consumes less than 550-nW power at 250-mV supply.

UR - http://www.scopus.com/inward/record.url?scp=84863905313&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84863905313&partnerID=8YFLogxK

U2 - 10.1145/2180878.2180882

DO - 10.1145/2180878.2180882

M3 - Article

AN - SCOPUS:84863905313

VL - 8

JO - ACM Journal on Emerging Technologies in Computing Systems

JF - ACM Journal on Emerging Technologies in Computing Systems

SN - 1550-4832

IS - 2

M1 - 10

ER -