We have previously presented a system for sensory feedback in hand prostheses to provide the user with a sense of touch and here we elaborate the details of the technical solution. The system aims at relocating tactile input from a hand prosthesis equipped with sensors to the forearm skin using actuators mounted on the forearm, thus providing sensory information to the user. The concept utilises the neural mechanoreceptors for pressure in the forearm skin. Every time the user touches and manipulates an object with the prosthesis, the mechanoreceptors of the forearm skin are activated by the tactile display.
One concern was that the actuator relaying pressure onto the forearm would generate artefacts in the recorded myoelectric signals as electrodes would also be located on the forearm. The EMG measurement from very closely located electrodes (1 cm) shows some influence on the recorded EMG that was easily filtered out by a high pass filter with a 20 Hz corner frequency. This filter would also reduce artefacts in the EMG originating from movement of the prosthesis. Filtering out signals below 20 Hz will also filter out EMG signals. However, the lower frequencies have been filtered out in earlier works of myoelectric control [25] where a high pass filter with a corner frequency as high as 200 Hz was reported with a high recognition ratio on a group of amputees. A cut-off frequency of 20 Hz should therefore not reduce the performance of a myoelectric control system.
An important issue is that the spatial resolution of the forearm is an order of magnitude less than the spatial resolution of the fingertips. In the work of Weinstein [26], two-point discrimination on different parts of the body have been investigated
The two-point discrimination of the forearm is about four cm suggesting this to be interdistance between actuators which closely corresponds with our placement. Thus the actuator elements need to be placed with a quite high degree of separation on the forearm. This put restrains on how well the sensory input could be fed back to the participants. The actuators were initially placed on a line except for the actuator for the thumb that was placed towards the hand. Testing different placements later revealed that a placement of the actuators reflecting the placements of the fingertips of an open hand drawn onto the forearm was more intuitive and also increased the distance between the actuators.
The proposed sensory feedback system provides dynamic and static pressure feedback to the user. This could also be used on patients that have undergone targeted muscle reinnervation, placing the tactile display on the reinnervated areas of the chest. However, this type of procedure is more suitable for a shoulder disarticulation amputation as to the invasive nature of the procedure.
The accuracy presented in our results with two individuals shows promising outcomes for the use of such a tactile display with five actuators to represent the five fingers. It is likely that an even higher accuracy would be gained after more training. Different methods of training could also help, e.g. reinforced learning were the participant is blindfolded and guessing which stimuli that have been applied where after the test supervisor provides feedback in the form of a correct answer. Having more sensing elements on the thenar and hypothenar regions of the hand seems to be a natural region to cover and may increase haptic perception. However, the result also shows that the five fingers were not easily separated by all participants and increasing the number of actuators would definitely demand more training of the user. The number of sensors and hence the number of actuators were based on having one sensing element per finger of the human hand.
The force level discrimination experiment had basically the same resolution as the finger discrimination experiment. Usually the force delivered by the actuator would be measured and controlled, however, using the displacement of the actuator lever will give a good force estimate.
A disadvantage with servomotors is their power consumption which is quite high. Power consumption will always be an issue in prosthetics and the usefulness of the proposed device as perceived by the user vs. power consumption are also fundamental parameters that needs to be investigated.
Actuators were mounted on the forearm on two participants providing sensory feedback addressing both spatial as well as the level of force of the stimuli. The evaluation of the system in the present study as well in the previously presented study where the system was applied on 11 healthy test subjects [21] suggests that this is a viable method for providing sensory feedback to forearm amputees using prosthetic devices.
A quite distal placement of actuators of the same size as the ones here proposed should be possible in a prosthetic socket without compromising the appearance of the socket. If the actuators were to be integrated into a socket, the socket itself would provide some attenuation of the sound generated by the motors. Using a sensory feedback system in a prosthetic device would also be to improve user acceptance of the prosthetic device as a whole.
Although it remains unrealistic to expect the proposed sensory feedback system to provide as accurate sensibility as a real hand, our system does provide a relatively simple and non-invasive way to restore rudimentary sensory feedback in prostheses used by transradial/-humeral amputees. Future work will focus on application of the system on amputees, training methods, cognitive aspects and finding a solution with actuators that have a lower power consumption.