| Popis: |
The aim of the investigation was to develop a control system combining command and proportional control of robotic devices based on electromyography (EMG) signals. Materials and Methods. EMG signals were recorded using 8-channel bracelet MYO Thalmic (Thalmic Labs, Canada). Command control of robotic devices was exercised by EMG patterns associated with 6 static hand gestures. The patterns were classified by periodic calculation of a root mean square value of an EMG signal for all channels with further recognition by a two-layer neural network based on back propagation algorithm. Proportional control was performed by calculating the mean absolute value of an EMG signal, and command execution speed adjustment proportional to this value. The software of the control unit was connected via wireless Bluetooth interface with a mobile robot assembled from a set of LEGO NXT Mindstorms (LEGO, Denmark). Results. We presented a soft and hardware platform combining command and proportional control of robotic devices based on EMG signals, and determined the parameters providing optimal characteristics of classification accuracy of EMG patterns, as well as the speed and accuracy of proportional control. We put forward and studied the following schemes combining command and proportional control: 1) the use of independent channels of different control types with recording EMG signals from both hands, 2) the use of independent channels of different control types with recording EMG signals from one hand only, 3) the use of all channels recording an EMG signal from one hand for classification and dynamic selection of a channel for proportional control, and 4) the use of all channels recording an EMG signal from one hand for classification with an average signal across all channels for proportional control. Conclusion. We proposed a novel system of combined command-proportional control of robotic devices based on the neuromuscular activity signals. We studies several schemes and chose the most preferable (Scheme 4) one, and found the optimal parameters for command classification accuracy, as well as speed and accuracy of proportional control. |
