Despite technological advances upper limb prosthesis, myoelectric control requires a long learning process. The absence of sensory feedback is very likely to impinge the appropriation of the prosthesis. We explored a sensory substitution alternative with various configurations and settings for vibrotactile feedback and myoelectric parameters.

Six vibrators were placed on a line or circumferentially around the arm. Space intervals between vibrators were absolute (2cm) or proportional to the length or the circumference of the arm. Dispositions as longitudinal proportional and absolute, circular proportional and absolute were tested. Sixty, 100 and 140ms of duration and intensities of 62.5; 100; and 167mA were explored. Estimate location of the vibration, and perceived intensity between 0 (no feeling) to 3 (strong) were assessed.

EMGs from biceps and triceps were recorded during isometric contraction to control the elbow velocity of an avatar displayed on a screen. EMGs were filtered and normalized from maximal contraction. A threshold of minimum muscle activity and a gain of the velocity control were adjusted to allow fast an intuitive control.

The circular proportional disposition elicited better discrimination results than the 3 others dispositions (P<0.05). Duration, intensity and disposition were all found to influence the success rate scores (3 ways ANOVA P<0.0001). Stimulations with small duration (60ms) were perceived as being produced with a lower level of intensity.

A threshold between 5 and 7.5% of the maximum force, and a velocity gain varying from 0.3 to 1.2rad.s-1 for a change in muscle contraction of 10% MVC were found to enable precise control of the avatar.

The circular proportional disposition of the vibrator is a well suited configuration for sensory substitution. Our next step will be to combine sensory feedback given by vibrator to the myoelectric control. This sensory substitution could improve prosthesis control, and may attenuate phantom limb pain.