Using Brain-Computer Interface to Control a Virtual Drone Using Non-Invasive Motor Imagery and Machine Learning

In recent years, the control of devices “by power mind” has become a very controversial topic but also been well researched in field state-of-the-art gadgets, such as smartphones, laptops, tablets and even smart TVs, medicine, to be used by people with disabilities for whom these technologies may only way communicate outside world. It is known that BCI skill can improved through practice training. This paper aims improve diversify signal processing methods implementation brain-computer interface (BCI) based on neurological phenomena recorded during motor tasks using imagery (MI). The aim research extract, select classify characteristics electroencephalogram (EEG) signals, which are sensorimotor rhythms, systems. article investigates systems interfaces, especially those use method acquisition MI tasks. purpose this allow users manipulate quadcopter virtual structures (external, robotic objects) simply brain activity, correlated certain mental undecimal transformation (UWT) reduce noise, Independent Component Analysis (ICA) together determination coefficient (r2) and, classification, hybrid neural network consisting Radial Basis Functions (RBF) multilayer perceptron–recurrent (MLP–RNN), obtaining classification accuracy 95.5%. Following tests performed, it stated biopotentials human–computer interfaces viable applications BCI. results presented show training produce rapid change behavioral performance cognitive properties. If more than one session used, beneficial increasing poor performance. To achieve goal, three steps were taken: understanding functioning involved; acquiring EEG signals rhythms tasks; applying optimizing extraction methods, selecting classifying neuronal networks.