In 1944, the hand surgeon Sterling Bunnell1 wrote, “Without sensation, a worker can scarcely pick up a small object, and he constantly drops things from his grasp. The so-called eyes of his fingers are blind” (p. 222). Users of electrically powered hand prostheses profit from an improved cosmesis and regain functionality, facilitating activities of daily life and interaction with society. However, their artificial hands suffer from a lack of adequate feedback compared with body-powered prostheses where information on both grasping forces and proprioception (e.g., hand posture) is obtained from the same harness and cables that are used to operate the hand. As a result, users of electrically powered hand prostheses control grasping of objects with the nonsensate hand by the use of vision. It is expected that the acceptance rate of prosthetic hands can be improved with a sensory feedback system.2 The importance of an adequate feedback system for controlling the grip force has been analyzed,3 and the need for implementation in prosthetic hands was identified as early as 1917 by Franz Rosset,4 who filed a patent on different mechanical methods, including a pneumatic feedback system. Since then, research groups all over the world have designed force and velocity feedback systems.5,6
Of the several feedback principles published and discussed extensively in the last decades, two are considered to have the best clinical viability: electrical surface stimulation (ESS) and mechanical surface stimulation (MSS)-like pressure or vibration.6 Direct stimulation of peripheral nerves might be another viable feedback method in the future, and worldwide research progresses rapidly. However, electrodes have to be implanted, and long-term functionality is one of the problems to be solved. The information capabilities of ESS and MSS have been compared7 with the conclusion being that both systems are approximately identical and as many as five stimulus levels can be discriminated reliably. With the advancements of electrically driven prosthetic hands, several ESS systems have been designed.8–12 However, problems associated with adaptation to the ESS13 and interference between the myoelectrical control system and the feedback signal leading to malfunction must be considered.7 Moreover, the range of signal current between the detection threshold (1 mA) and pain threshold (2 mA) is narrow, and pain sensations are reported to be the major disadvantage of electrotactile stimulation.7,13 For this reason, other researchers proposed vibration10,14,15 or pressure16,17 to transmit information. An advantage associated with mechanical stimulation for force feedback is the universal psychological acceptance.7,16 However, the actuators used for mechanical feedback have been too bulky and could not be integrated in a prosthetic hand. In the last decade, costs of small low-power pager motors have dropped significantly with widespread application in mobile phones. In prior studies, different types of these motors were evaluated successfully for application in shoulder pad displays18 and prosthetic feedback systems.19 In addition to the feedback actuator, an adequate sensor and electronics are needed for closed-loop control. Scott6 pointed out that there is a lack of viable sensors to detect the grip force within the prosthetic hand, especially when a cosmetic glove covers the force sensor. The goal of this study was to design and evaluate a force feedback system based on vibrotactile stimulation considering these criteria for a viable prosthetic sensory feedback system: physiological feedback signal, not felt as artificial, unpleasant, or painful; compact design; low cost; imperceptible to others (e.g., no buzzing sound); and low energy consumption.
DESIGN OF THE FORCE FEEDBACK SYSTEM
The new feedback system can be subdivided into three components (Figure 1):
1) A force sensor (“FS” in Figure 1) based on a widely used and inexpensive standard conductive polymer pressure sensor with an outer diameter of 8 mm (FSR® Type 149, Interlink Electronics Inc., Camarillo, CA). These sensors are made of three coated plastic film layers, a semiconducting bottom layer, a top layer with two interdigited conducting structures printed on them, and a spacer layer that separates these two layers. The sensor has a logarithmic resistance-versus-force characteristic that typically decreases from approximately 2 MΩ to 10 KΩ when pressure is applied. To achieve a force-to-pressure transduction, the sensor was modified by mounting a silicone rubber structure with the shape of an inverted dome onto it. In this way, accurate measurement of forces is ensured20 and the sensor is mechanically protected against sharp edges. The force sensor was connected to the palmar tip of the prosthesis’s aluminum thumb “bone.”
2) A vibrotactile stimulation device (“VM” in Figure 1) consists of a coin-type vibration motor (Sanko Electric Type 1E120, Jameco Electronics, Belmont, CA) that is typically used as a pager motor in mobile phones. Both vibrational frequency and amplitude are linked and increase with the voltage applied to the motor. It is characterized by the following technical data: Dimensions, 14 mm (diameter) × 4 mm (height); operating voltage range, 1.5 to 4 V; and a maximum current draw of 40 mA.
3) The integrated electronic board (“EB” in Figure 1) converts the variable resistance of the force sensor into an electrical signal to operate the vibration motor. It is designed as simply as possible but also as functional as needed, with only low-cost standard electronic components being applied. The design combines a voltage divider setup, as shown in Figure 2, with a potentiometer to adjust the threshold level of feedback. Because of the highly resistant input signal, an operational amplifier is used in a noninverted mode and a driver transistor supplies the vibration motor with a voltage in the range of 0 to 3.5 V. Low forces applied to the force sensor result in a vibration motor frequency of about 50 Hz, whereas high forces produce a frequency of about 80 Hz. In a former experiment it was demonstrated that four to six stimulation levels can be identified reliably in this frequency range.19 The technical data of the system are: operation voltage, 6 to 12 V; noise level at 30 cm distance, 38 to 41 dB; power consumption, 24 to 270 mW. The force feedback system was integrated in a myoelectrically controllable hand prosthesis (Dynamic Mode Control Plus System Electric Hand, No. 8E38 = 6, Size 7¾, Otto Bock, Duderstadt, Germany).
Clinical evaluation was aimed at demonstrating the effectiveness of the force feedback system in myoelectric prosthetic hand operation without the help of vision. For this purpose, it was analyzed whether the applied forces correlated with the objects’ weights and also whether the requisite grasping force was reduced significantly by the use of a feedback system.
Five habitual myoelectric prosthesis users (4 men, 1 woman, aged 22–38 years) with left hand transradial amputation took part in the experiment after having signed an informed consent form. Their task was to grasp, lift, and hold a hand dynamometer (MLT003, ADInstruments Pty Ltd, Castle Hill, Australia) with a five-finger pinch grip without force feedback. Four different weights were attached to the dynamometer in a random order. These weights were invisible to the test persons, as they were kept in a box below the handle of the dynamometer (Figure 3). Measurements were repeated until data of each weight level were recorded four times with the modified prosthesis with deactivated feedback system. The first measurements were followed by a 5-minute break and another 5-minute training period to familiarize the testers with the feedback signal. The force feedback system was activated with the vibration motor being placed between the skin and the silicone liner of the prosthesis in the dorsal region of the residual upper limb. First, the vibration threshold was adjusted to each subject individually so that grasping the dynamometer with a pinch force of 10 N corresponded to a vibration signal that was subjectively perceived as a “weak” signal and a force of 60 N as a “strong” vibration. Then, the test persons were asked to grasp and hold the dynamometer with different weights attached to it. They were told the amount of weight prior to grasping it. In a second test series, the same dynamometer-weight combinations as in the first series had to be grasped and held for several seconds with an activated feedback system. In a third test series, a vibration motor was activated that was attached to the hand mechanics, instead of the silicone liner. At the end of data acquisition, the test subjects were asked whether vibration was unpleasant and to comment on the usefulness of this feedback system.
The amplified signal from the strain-gauged hand dynamometer was recorded and stored on a PC with a data acquisition card (PCI-6024E; National Instruments Corporation, Austin, TX). Using MatLAB software (National Instruments Corp.), a linear correlation curve for the sensor output was calculated in the range between 0 and 130 N. First trials revealed that the curved profile of the dynamometer handle was difficult to grasp with the hand prosthesis. Thus, two adaptors made of acrylic glass with one flat surface and one curved shape were designed and connected to levers of the handle. To reduce the influence of the grasping position in relation to the handle, the target position for the thumb was marked on one lever. The weight of the dynamometer handle can be varied on five discrete levels from 403 g with no additional weight attached to the handle to 1,412 g.
Mean values were calculated from the four values measured for each weight level in the four test series. Then, linear regression curves and the corresponding coefficients of determination were calculated for the test series. Finally, the levels of significance of differences between the exerted forces with and without closed-loop feedback were calculated from the mean values of the five test persons using the non-parametric distribution-free Mann-Whitney test (U test).
RESULTS AND DISCUSSION
The mean forces generated by the myoelectric hand with and without feedback that were measured for all weight levels of the dynamometer are depicted in Figure 4. It can be seen that the requisite grasping forces were lower when feedback was available. The forces were found to decrease by 16% to 76%, with a mean value of 37% when vibration was applied indirectly to the hand. A wider difference of 30% to 77% with a mean decrease of 54% was found when feedback was applied directly to the skin of the residual limb. In both cases of force feedback, the decreases were significant (p < 0.01, two-tailed U test). The significant reduction of the requisite grasping force in our study closely correlates with that obtained in the work of Lundborg et al.12 Consequently, the use of a closed-loop feedback system based on vibration or electrical stimuli can be recommended. Figure 4 also indicates that the grasping of higher loads tends to correlate with higher forces applied by the prosthetic hand. Highest linear correlation with the object load was found when feedback was applied directly to the skin, which is reflected by a coefficient of determination of r2 = 0.9729. This approximates best the characteristic of natural hands, where a strong linear correlation (r2 = 0.9992) was found by measurements in our lab (unpublished results). No correlation between the sum of forces and the percentage reduction of forces by applying feedback was found. In the subject who exerted the highest absolute forces, a medium percentage reduction was found. However, the subject with medium grasping forces exhibited the highest percentage reduction of forces. Despite the better percentage force reduction of a direct feedback signal (stimulation device on the skin), this arrangement has two drawbacks. First, wiring between the vibration motor and the electronic board crosses the wrist, which may be a source of failure if the wrist can be rotated. Second, the application is restricted to a socket system with silicone liner technology.
The components used in the experiment are characterized by low weight and compact size, thus fulfilling the design constraints of a prosthetic feedback system. In addition, they are inaudible, especially when typical office or public background noise is present. The vibration frequencies used in this experiment are lower than the frequency of highest sensibility for cutaneous vibrational perception, which is approximately 250 Hz,6,21 and the number of discernible stimulation levels would be higher in an expanded frequency range of stimulation. However, according to the results of Sueda and Tamura,15 the use of vibration frequencies below 100 Hz decreases the level of noise perception because the auditory sensibility also decreases from 250 to 50 Hz. Compared with other sensor principles such as strain gauges used in experimental prosthetic hands,11,16,22 the costs are significantly lower and the electronics for signal processing considerably easier.
Shannon7 and Verillo21 point out that a variation of the contact pressure leads to inconsistent detection sensitivity. When designing the new feedback system, this point was considered by embedding the vibration motor in the silicone liner or by attaching it to the base plate of the mechanical hand. Thus, the contact pressure of the stimulation device was maintained steady in both arrangements.
Acceptance of the feedback stimulus by all test persons was good when vibration was generated indirectly at the mechanical hand. Furthermore, the stimulus was accepted by four of the five subjects when applied directly to the residual forelimb; only one female test person disliked this location for feedback stimulation. After feedback had been turned off, three subjects reported that they felt “something was missing.” With closed-loop feedback prosthetic hands, users reported a sense of confidence provided by the sensory information. This confirms the statements made by Shannon.22 Two of the test persons remarked that the vibration signal becomes more disturbing than helpful after some seconds. Consequently, we suggest limiting the feedback signal to a definite period of 3 to 5 seconds. Thus, a modified prosthetic force feedback system with an additional timer on the electronic board was developed and will be used for additional experiments. Two positive effects of a feedback system with time were also identified: a reduced power consumption of the feedback system and a reduced influence on the skin’s vibration detection capability.9
A comprehensive review of the need for feedback in myoelectric prostheses is presented. Of different methods, vibration feedback was preferred because of the availability of very compact inexpensive vibration motors. A new force feedback system is presented that complies with general design constraints for prosthetic hands. It is characterized by a low weight, low costs, low power consumption, and compactness. The acceptance of the vibration feedback signal was good, especially when the stimulus was applied indirectly to the prosthetic hand. A significant reduction of grasping force was found when using the feedback system. It is expected that grasping with a myoelectric prosthesis will become more precise and confident when a feedback system is used.
The authors gratefully acknowledge the valuable help of Mr. Joachim Frühauf and the staff of the orthopedic workshop, Orthopedic University Hospital, Heidelberg, and that of the staff of Pohlig Orthopädietechnik, Tübingen.
1. Bunnell S. Surgery of the Hand.
Philadelphia: Lippincott; 1944.
2. Lundborg G, Rosén B. Sensory substitution in prosthetics. Hand Clinics
3. Westling G, Johansson RS. Factors influencing the force control during precision grip. Exp Brain Res
4. Rosset F. Künstliches Glied. German patent application. No. DE301108, 1917.
5. Childress DS. Closed-loop control in prosthetic systems: historical perspective. Ann Biomed Eng
6. Scott RN. Feedback in myoelectric prostheses. Clin Orthop Rel Res
7. Shannon GF. A comparison of alternative means of providing sensory feedback on upper limb prostheses. Med Biol Eng
8. Kato I, Yamakawa S, Ichikawa K, Sano M. Multifunctional myoelectric hand prosthesis with pressure sensory feedback system, Waseda Hand 4P. Proceedings of the 3rd International Symposium on External Control of Human Extremities, Dubrovnik, Yugoslavia, August 25–30, 1969. In: Advances in External Control on Human Extremities
, Yugoslav Committee for Electronics and Automation (ETAN), Belgrade, Yugoslavia, 1970:155–170.
9. Schmidl H. The importance of information feedback in prostheses for the upper limbs. Prosthet Orthot Int
10. Shannon GF. Characteristics of a transducer for tactile displays. Biomed Eng
11. Wang G, Zhang X, Zhang J, Gruver WA. Gripping force sensory feedback for a myoelectrically controlled forearm prosthesis. Proceedings of the IEEE International Conference on Systems, Man, and Cybernetics, Vancouver, Canada, October 22–25, 1995, pp 501–504.
12. Lundborg G, Rosen B, Lindström K, Lindberg S. Artificial sensibility based on the use of piezoresistive sensors. J Hand Surg (Brit/Europ)
13. Kaczmarek KA, Webster JG, Bach-y-Rita P, Tompkins WJ. Electrotactile and vibrotactile displays for sensory substitution systems. IEEE Trans Biomed Eng
14. Conzelmann JE, Ellis HB, O’Brian CW. Prosthetic device sensory attachment. US Patent application US2656545, 1953.
15. Sueda O, Tamura H. Sensory device for the artificial arm. Paper presented at the 8th International Conference on Medical and Biological Engineering, Chicago, Illinois, July 21–25, 1969.
16. Patterson PE, Katz JA. Design and evaluation of a sensory feedback system that provides grasping pressure in a myoelectric hand. J Rehabil Res Dev
17. Sears HH, Iversen EK, Hays KB, Dyck AD. Method and apparatus for controlling an externally powered prosthesis. US Patent application US5888213, 1999.
18. Toney A, Dunne L, Thomas BH, Ashdown SP. A shoulder pad insert vibrotactile display. Proceedings of the 7th IEEE International Symposium on Wearable Computers, White Plains, New York, October 21–23, 2003, pp 35–44.
19. Pylatiuk C, Kargov A, Schulz S, Bretthauer G. A sensory feedback system for prosthetic hands. Biomed Eng
20. Kargov A, Pylatiuk C, Martin J, Schulz S. Determination of the grip force distribution in functional grasping. Technol Health Care
21. Verillo RT. Effect of contact or area on the vibrotactile threshold. J Acoust Soc Am
22. Shannon GF. A myoelectrically-controlled prosthesis with sensory feedback. Med Biol Eng Comput