Hand Detection for Gesture Recognition

Gesture recognition plays an essential role in a number of fields, facilitating natural and intuitive interactions between humans and machines. It helps people with physical disabilities to interact with technology using gestures, offering an alternative to the traditional input methods. It can also be considered as a communication tool for sign language translation. In this paper we propose an implementation of gesture recognition system using an acquired dataset under uniform background. The system utilizes various machine and deep learning methods, including a skin detection segmentation, the Local Binary Pattern (LBP) descriptor characterization, followed by K-Nearest Neighbor (KNN) classification. Additionally, a Convolutional Neural Network (CNN) was trained and implemented. The results were highly satisfactory, achieving 97.25% of accuracy using the KNN on segmented data.