Affordable Real-Time Hand Gesture Detection Using Random Forest

This paper represents a hand gesture recognition system that classifies each gesture from a given gesture alphabet, makes use of the Random Forest Algorithm alongside with Pickle library to efficiently save and train the model. This system is capable of identifying all the alphabets along the continuous stream of video. The approach focuses on training the Random Forest model to identify between the gesture and non-gesture instances based on the features extracted from the training dataset. The saved models seamlessly allow for integration for the real-time use. The results produced truly shows the efficiency of the Random Forest approach achieving the desired accuracy without having or requiring any additional complex preprocessing or additional spatial information. We conclude the advantage of the approach as to how low the computational cost can be attained and ease of implementation, while keeping in mind the area for enhancement for future perspective.