New Results on Classifying EMG Signals for Interfacing Patients and Mechanical Devices

In modern days the goal of rehabilitative robotics is to take advantage of robotics-inspired solutions in order to assist people affected by disabilities using physical training assisted by robots. In this way the rehabilitative exercises could be autonomously performed by the patients, with a reduced involvement of the therapist, making high-intensity rehabilitative therapy an affordable reality. Moreover high-precision sensors integrated in rehabilitation devices would allow a quantitative evaluation of the progresses obtained, effectively comparing different training strategies. That would represent a huge scientific achievement in a field where evaluations up to this day are performed only by means of subjective observations. Important results were obtained in rehabilitative robotics, but results in the field of the hand rehabilitation are poorer, due to the high complexity and dexterity of the organ. This chapter proposes to integrate the detection of the muscular activity in the rehabilitation loop. A new EMG analysis tool was developed to achieve a reliable early recognition of the movement. Experimental results confirmed that our system is able to recognize the performed movement and generate the first control variable after 200 ms, below the commonly accepted delay of 300 ms for interactive applications. This shows that it is possible to effectively use an EMG classifier to obtain a reliable controller for a flexible device, able to assist the patient only after having detected his effort.