In the first mock app we made, we made use of the OpenCVs deep learning libraries to capture the blink of a person , which is then processed to play or stop the music. The webcam feed was captured live via the app and using functions in the libraries, it was detected whether or not the user was blinking. Successful registration of a blink was then used to play or pause the music in the mock app. Using only a single blink as a criteria had its clear problems since sometimes the program picked up unintentional blinks. The criteria was then changed to two consecutive blinks to play or pause the music for better accuracy.
Using Webcam to capture the hand gestures of the subject. Using OpenCv’s Computational Abilities and deep learning algorithms to fine tune a model which can predict the hand gesture of a person, and execute an action accordingly. Detecting the subject’s gesture using the created model. Using the detected hand-gesture to interface with computer.
Average Accuracy for blink detection: 60.66% Average Accuracy for Hand Gesture Detection: 80% Average accuracy for viewing distance(blink-detection): 67% Average accuracy for viewing distance(hand-gesture): 83% Average accuracy for viewing distance(blink-detection): 67% Average accuracy for far distance(blink-detection): 35% Average accuracy for far distance(hand-gesture): 45%