Hand Gesture Recognition System using Feedback CNN

Sign language recognition is a pivotal challenge in bridging the interaction gaps for the hearing impaired. The research addresses this challenge by proposing an innovative system employing computer vision and methodologies to enhance the efficiency of sign language interpretation. The primary goal is to recognizing diverse sign gestures. Leveraging OpenCV for initial preprocessing tasks such as hand detection and gesture segmentation, the system employs advanced techniques like Histogram Oriented Transitions (HOG) and feedback Convolutional Neural Networks (CNNs) for feature extraction. Instead of traditional CNN methods, this technique integrates with Feedback mechanism where the user is prompted with feedback if confidence threshold goes below 0.5. This feedback-based approach demonstrates significant superiority over individual CNN feature extraction techniques, indicating its potential in significantly improving communication accessibility for the hearing-impaired community.