Hand Gestures Recognition using Inertial Sensors Through Deep Learning
Dhawal Mali, Atul Kamble, Shubham Gogate, Jignesh Sisodia · 2021
Gestures provide intuitive ways to facilitate interaction between humans and computational systems. Current gesture recognition methods use multi-step processes wherein the data stream goes through multiple processes like pre-processing, segmentation, decision-making, etc. We propose an end to end gesture recognition system using DeepLearning \& TransferLearning models where the stream of data from Microelectromechanical systems(MEMS) consisting of a three-axis accelerometer data stream is used to recognize the gestures in real-time. Gestures are communicated wirelessly to Personal Computers using Bluetooth for control purposes. Users can incorporate their gestures into a collection of pre-existing gestures to make a truly customizable system. To test the setup, a general-purpose desktop application(Action Mapping Interface) is implemented which simulates the keyboard and mouse events \& runs complex commands using gestures identified by the neural network.