Light-weight single trial EEG signal processing algorithms: Computational profiling for low power design

Brain Computer Interface (BCI) systems translate brain rhythms into signals comprehensible by computers. BCI has numerous applications in the clinical domain, the computer gaming, and the military. Real-time analysis of single trial brain signals is a challenging task, due to the low SNR of the inco...

Full description

Saved in:
Bibliographic Details
Published in:2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society Vol. 2011; pp. 4426 - 4430
Main Authors: Ahmadi, Ali, Jafari, Roozbeh, Hart, John
Format: Conference Proceeding Journal Article
Language:English
Published: United States IEEE 01.01.2011
Subjects:
Accuracy
Algorithms
Brain - physiology
Classification algorithms
Electroencephalography
Electroencephalography - methods
Feature extraction
Fourier Analysis
Humans
Man-Machine Systems
Signal processing
Signal processing algorithms
Signal Processing, Computer-Assisted
Support Vector Machine
Support vector machines
ISBN:9781424441211, 1424441218
ISSN:1094-687X, 1557-170X, 2694-0604, 2694-0604
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Brain Computer Interface (BCI) systems translate brain rhythms into signals comprehensible by computers. BCI has numerous applications in the clinical domain, the computer gaming, and the military. Real-time analysis of single trial brain signals is a challenging task, due to the low SNR of the incoming signals, added noise due to muscle artifacts, and trial-to-trial variability. In this work we present a computationally lightweight classification method based on several time and frequency domain features. After preprocessing and filtering, wavelet transform and Short Time Fourier Transform (STFT) are used for feature extraction. Feature vectors which are extracted from θ and α frequency bands are classified using a Support Vector Machine (SVM) classifier. EEG data were recorded from 64 electrodes during three different Go/NoGo tasks. We achieved 91% classification accuracy for two-class discrimination. The high recognition rate and low computational complexity makes this approach a promising method for a BCI system running on wearable and mobile devices. Computational profiling shows that this method is suitable for real time signal processing implementation.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISBN:9781424441211
1424441218
ISSN:1094-687X
1557-170X
2694-0604
2694-0604
DOI:10.1109/IEMBS.2011.6091098