Human-Machine Interfaces in rehabilitation engineering often use activity signals. Examples are electrical wheelchairs or prostheses controlled by means of muscle contractions. Activity signals are user-dependent and often reflect neurological weaknesses. Thus, not all users are able to operate the same control scheme in a robust manner. To avoid under- and overstraining, the interface ideally uses a control scheme which reflects the user’s control ability best. Therefore, we explored typical phenomena of activation signals. We derive criteria to quantify the user’s performance and abilities and present a routine which automatically selects and adapts one of three control schemes being best suited.