Thought Scope: Human Cognitive Thinking Classification from Eye Position using Deep Learning and Digital Image Processing

Human eye serves as a unique gateway to understand thoughts and cognitive activities. As the eyes always convey the truth, throughout history, we’ve been able to perceive a person’s emotions and sentiments to a certain extent by observing their eyes. With the dynamic technological advancement in the world, we want to level up from just reading emotions from the eyes to deciphering what that person might be thinking. Understanding the potential thoughts of individuals based on subtle change of eye positions, can help us immensely in several sectors. In this paper, we are proposing an approach to decode human cognitive processes by analyzing eye images captured from distinct eye positions. Our system uses captured images of eyes, then it utilizes digital image processing and Convolutional Neural Network (CNN) to detect the eye position. It compares the position to seven predefined classes and determines the closest match. Based on the similarity of the eye positions, our system will provide textual feedback about what the person might be thinking. To classify cognitive thinking based on eye position, CNN was used with transfer learning from two pretrained models- MobileNetV2 and NasNetLarge, where MobileNetV2 obtained the highest accuracy of 98.65% and is capable of delivering accurate textual representations of the potential thoughts remarkably.