Brain-machine interfaces (BMI) have been considered since many years as the most promising approach to the palliation of severe motor handicap. This thesis describes a hybrid brain-machine interface, designed specifically for patients suffering from Duchenne muscular dystrophy. Our hybrid BMI uses signals recorded by electroencephalography (EEG), electromyography (EMG), and joystick sensors. Signal processing enables the hybrid BMI to detect a movement or movement intent at different levels of the motor command chain. Joysticks are used as long as the patient is able to activate them, then when motricity deteriorates with the disease evolution, the hybrid BMI takes EMG signals into account and finally EEG signals. We have developed an original method for processing EEG signals, allowing the system to select features that a human expert considers as the most discriminant. Performance has been assessed on a data set used as a reference in the BMI community, as well as on data that we have recorded from healthy subjects in our laboratory. Our hybrid BMI controls the trajectory of a moving object – either real or virtual – through three actions, corresponding to a movement or an intent of movement of the right hand, the left hand, or both hands simultaneously. An additional degree of freedom can be considered by integrating the detection of attempted feet movements.
Directeurs de thèse : Rouillard José et Cabestaing François Rapporteurs : Louis-Dorr Valérie et Bourhis Guy Examinateurs : Tiffreau Vincent et Tarpin-Bernard Franck
Thesis of the team BCI defended on 01/12/2016