Hand Tracking System

Supervised by Prof.Khalaj, System Security and Edge Computing Lab, Sharif University of Technology

In the first step, I studied recent models such as mediapipe’s hand solution. I tested this framework on different devices like Raspberry Pi4 and Android mobile phones with both python and cpp. Following that, I aimed to enhance this simple hand tracker by adding some features. I implemented a hand gesture recognizer that utilizes 21 calculated key points from mediapipe’s model and classifies the gestures using an MLP network. Additionally, I incorporated a movement recognizer into our solution. The model records the points’ history and passes them to an MLP/LSTM network to predict the action. This solution can be used for specific or general purposes, such as in vehicles as a remote demo