Hand gesture recognition based on Time-of-Flight sensors

With the rapid development of technology, hand gesture recognition has drawn wide attention due to its versatility and efficiency. Additionally, it is an intuitive way of more natural human-machine interaction. Over the past few years, gesture recognition has made its way in applications varying from sign language to medical assistant to the automotive industry. Moreover, the COVID-19 pandemic has accelerated the demand for touchless technology in public facilities and sanitary equipment. The majority of the research has been devoted to vision-based gesture recognition techniques. Yet, they present significant constraints: Reduced performance in lower visibility conditions and the need for high computational power. Based on the Gesture Recognition and Touchless Sensing Market Global Analysis, Sensor-based technology is anticipated to endure the largest share of the market during the next decade. Two types of sensors are majorly utilized in touchless sensing devices: infrared sensors and capacitive sensors. However, each of these sensors has proven to have significant drawbacks. In this dissertation, a design of a touchless sensor prototype has been proposed by using time of flight technology. The strategy aims to obtain the best configuration while maintaining a low cost and reliability. The time-of-flight principle is based on the speed of the light. An emitter sends photons that are reflected by the target and detected by the receiver (called SPAD for Single Photon Avalanche Diode). The time difference between the emission and the reception provides the actual distance of the target in millimeters with high accuracy. The proposed system consists of a horizontal array of three time-of-flight sensors. These distance measurements of these sensors are used for data collection from seven hand gestures (up, down, left and right, clockwise (CW), counterclockwise (CCW), unknown) performed by the human at a specified distance from the sensor’s prototype. Afterward, the hand gestures are then classified using Deep Artificial Neural Network. The time-of-flight sensors offer high accuracy of 94%, and it was perform verified by a Graphical user interface designed as a car dashboard. Furthermore, this dissertation provides a perception of other sensing technologies through a comparative performance evaluation of the well-known types of gesture sensors available in the market.