Dr. Mohsen Iftikhar-4-confx

skirlorangeΒιοτεχνολογία

1 Οκτ 2013 (πριν από 3 χρόνια και 8 μήνες)

95 εμφανίσεις



Title

Suggested Hybrid Approach for Robust Classification of EEG Data for Brain
Computer Interface,

Author
-
s

Ijaz Ali Shoukat, M. Iftikhar,

Contact lnfo

mohsinunsw@gmail.com

Department

Computer
science

Major


citation

In Proceedings of
WORLDCOMP’10
-
International Conference on Bioinformatics
Computational Biology, BIOCOMP 2010, July 12
-
15, 2010, Las Vegas Nevada,
USA, 2 Volumes 2010. (it is top ranked conf. in Microsoft Academic Search),
2010

Year of
Publication

2010

Publisher

International Conference on Bioinformatics Computational Biology

Sponsor


Type of
Publication

Conference

ISSN


URI/DOI


Full Text
(Yes,No)


Key words

EEG based BCI, SVM, HMM, LDA, MLP, Feature Extraction
Methods

Abstract

The latest inclination of classifying the Electroencephalographic dataset using machine
learning methods has an effectual implication in Brain Computer Interfacing Systems. The
precedent studies have proposed diversified mechanisms for recording the human
brain
activity for generating control signals using taxonomical techniques. Moreover, efficacious
and optimized ways to preprocess the signals and to learn hidden patterns for evolving a
classifier have been put forward. Designing of such systems requires,

handling two most of
the influential perspectives in BCI Framework i.e. signal preprocessing and classifier
learning. The prior studies have itemized a number of tactics to resolve issues in signal
preprocessing (particularly noise and artifacts removal),

preparation of training set and
classifier learning. Unfortunately, the accessible methods have limited efficiency,
prediction accuracy and longer training time due to less deep comparative analysis for
classification tasks. Therefore, systems are develop
ed by using empirical methods by
considering limitations on theoretical models. This paper analyzes and discusses these
limitations and challenges in order to suggest a new combined approach of Brain
Computer Interface (BCI) classifiers. The exhibited rese
arch in this paper emphasizes on
the need of the development of more robust, adaptive and noise tolerant machine learning
schemes for the development of such systems. Furthermore, this survey proposes the
implementation of Common Spatial Pattern (CSP) meth
od for feature extraction with
stacked ensemble learning technique to elaborate systematic and authentic classifier to
generate control signals in order to regulate the external devices interfaced with human
brain. Hence, the present work brings novel cont
ribution to machine learning techniques
by overcoming the limitations of available methods and through introducing new robust
and effective means.