Facial Gesture Recognition for Drone Control

In recent years personal unmanned aerial systems (UAS) or drones have become part of everyday life. In this paper the design and implementation of an intelligent drone controller system that can be used for drone flight navigation in real time through recognition of facial gestures is presented. To do this, the drone’s pilot wears an Electroencephalogram (EEG) headband containing three silver/silver chloride (Ag-AgCl) electrodes placed on the pilot’s frontal cortex. The drone commands implemented in this project were: move up, move down, and move forward and backwards. The corresponding facial gestures for these commands were: raising eyebrows, hard eye blinking, clenching the user’s teeth, and rest position. The EEG signals associated with each facial expression were classified via an artificial neural network (ANN). To acquire the EEG signals, the Cyton board (8-channel biosensing board from OpenBCI) was used. Matlab software was used for denoising and feature extraction of the EEG signals, and for the design and training of the ANN. To be able to classify the EEG signals, feature extraction methods were performed. For the feature extraction, three statistical quantities were computed from each facial gesture collected from the subjects. The statistical parameters were: the standard deviation, root mean square, and mode. Then, the Neural Net Pattern Recognition tool from Matlab was used for the implementation and training of the ANN. After the ANN model was created, the output of the ANN was used to control a small drone. Results of the ANN training yielded a 95.5% accuracy in the classification of the facial gestures. Finally, the intelligent drone controller system was tested with the drone in real time, proving that our initial goal was met.