Real Time Gesture Recognition

2 minutes

December 1, 2022

Electromyography (EMG) sensors collect muscle actvity which can be used for accurately predicting hand gestures. With a custom designed software, training data can be collected from human participants as well as real time testing can be performed with a trained model.

Check next project: Accurate detection of 16 hand gestures

For the gesture recognition project, EMG sensors were placed on the user’s wrist and a custom designed software was used to perform the real time gesture recognition. MATLAB GUI was used to design the user interface as the EMG manufacturer provides ready-to-use codes for MATLAB. The custom designed software had the following features:

Customize EMG device: Add any EMG device to the interface, with changeable channel numbers, sampling frequency, gain, bandpass filtering.

Training Mode: List of commonly used hand gestures, with options to set the number of repetitions, contraction time, rest time etc.

Testing Mode: Real time prediction where the user performs the gesture and the system predicts it using machine learning models. The UI can customize the machine learning model used.

Database Development: The software allows the user to register the training data and load them in future testing sessions.

Significance

Gesture recognition has been predominantly used for control of prosthetic arms. In the present scenarios, it can be used in home applications such as turning off lights, changing TV channels, etc. It can also be used in VR games where gesture control can be implemented.

Some functionalities of the the softwares are shown below: