Biometric scanners, such as face or finger-print recognition, have become increasingly less secure. With these technologies struggling to maintain the needed amount of security people desire, new biometric technologies are needed. EEG recognition is much harder to fake, which is why it could be used for secure biometric identification.
Our primary target contribution is to develop image classification models to accurately classify EEG signals. Second, we would like to evaluate the efficacy of our model by testing it on a dataset of 105 individuals.
We successfully built a logistic regression model to perform classification at an accuracy of 78%. Additionally, we were able to create a Feed Forward Network to classify at an even higher accuracy of 94%. More interestingly, we find that CNNs struggle to perform this classification with an accuracy of only 71%.
This project was developed for CIS519: Applied Machine Learning with contributors Harsha Santhanam and Sandhya Bellary. Our project mentor was Ricardo. I primarily worked on data extraction, data cleaning, Feed Forward Network and the Convolutional Neural Network.