Text

theses

Optical brain imaging using functional near infrared spectroscopy (fNIRS) offers a non-invasive imaging tool for monitoring brain activity. fNIRS is a safe imaging technique offering high temporal resolution, making it an attractive choice for brain-computer interfaces, as well as for real-time and long-term monitoring of brain function. While often lauded for its portability, the application of fNIRS has been mostly limited to laboratory environments. In order to improve the portability and wearability of fNIRS systems, it would be desirable to reduce the number of optodes, without compromising the performance. Using a large of number of optodes, while resulting in better coverage, increases the setup time, causes added discomfort, and hence, reduces the duration of wear time. This work focuses on the application of fNIRS systems for decoding brain states and brain computer interfaces, and considers the problem of two-class classification in the case where the number of optodes are limited to those covering only the prefrontal cortex. Pairwise functional connectivity for various time windows have been computed as features. Extensive classification experiments have been performed using Support Vector Machines (SVM) and Convolutional Neural Networks (CNN) as classifiers. Results obtained from 8 subjects' data, indicate that it is feasible to create predictive models based on fNIRS-based functional connectivity measures of prefrontal cortex.

ETD_8672

M.S.

Includes bibliographical references

by Anthony Yang

doi:10.7282/T34J0JBZ

ETD graduate

The author owns the copyright to this work.