Biometric Based Secure Machine Authentication Method Using Hybrid Learning
Journal of Industrial Engineering Research, Vol. 1(6), Pages: 1-6, September 2015
6 Pages Posted: 13 Jul 2017
Date Written: June 16, 2015
Abstract
Systematic authentication and usage of several other computerized machines is a great problem for patients having motor rehabilitation due to stroke or spinal injuries. Brain Computer Interface (BCI) technology provides an alternative heuristic of controlling the machines with just neural activity without imposing any physical effort. The recording of neural activity (brain biometric signals) associates several learning issues such as outlier, non-stationary patterns, multi-dimensions, variable training sizes, timing variations, noise and artifacts. Due to these issues, learning and classification provide inadequate results under any single classifier. Hybrid training and classification methods can overcome these issues but the question of performance and feasibility matters is still open in this case. Several hybrid classifiers have been analytically reviewed and evaluated in this article to propose optimal hybrid learning and classification method for BCI based medical or security systems. The analysis shows that mutual assembly of HMM-SVM-LDA under CSP is an optimum approach to escape majority of learning issues. The presented analysis is not only significant for the development of BCI based security systems but it is also important for medical applications
Keywords: Biometric, Electroencephalography, Brain-Computer-Interface, Learning, Classification
Suggested Citation: Suggested Citation