Vehicle Control Using Hand Gestures

Abstract: Accurate gesture recognition in real time is a very challenging problem. If an architecture that has the capacity of correctly recognizing gestures based on context, can adaptively learn new gestures taking into account variations of the same gesture is built, it can be used to solve a multitude of problems from sign language translation, detection of body language and micro expressions to providing an intuitive method for control of automobiles. The problem can be solved with coloured gloves and markers; however, these models do not generalize well. Boundary based gesture recognition has also been carried out, however these models consume a lot of time during prediction and are complicated to build. This project tries to solve both these problems by the construction of a simple machine learning architecture for real time gesture recognition that also generalizes well.

Keywords: Image processing, sequence learning, adaptive learning, machine learning, hand gesture detection, hand motion detection.