A compact solution for vibrotactile proprioceptive feedback of wrist rotation and hand aperture

Abstract

Background: Closing the control loop between users and their prostheses by providing artificial sensory feedback is a fundamental step toward the full restoration of lost sensory-motor functions.

Methods: We propose a novel approach to provide artificial proprioceptive feedback about two degrees of freedom using a single array of 8 vibration motors (compact solution). The performance afforded by the novel method during an online closed-loop control task was compared to that achieved using the conventional approach, in which the same information was conveyed using two arrays of 8 and 4 vibromotors (one array per degree of freedom), respectively. The new method employed Gaussian interpolation to modulate the intensity profile across a single array of vibration motors (compact feedback) to convey wrist rotation and hand aperture by adjusting the mean and standard deviation of the Gaussian, respectively. Ten able-bodied participants and four transradial amputees performed a target achievement control test by utilizing pattern recognition with compact and conventional vibrotactile feedback to control the Hannes prosthetic hand (test conditions). A second group of ten able-bodied participants performed the same experiment in control conditions with visual and auditory feedback as well as no-feedback.

Results: Conventional and compact approaches resulted in similar positioning accuracy, time and path efficiency, and total trial time. The comparison with control condition revealed that vibrational feedback was intuitive and useful, but also underlined the power of incidental feedback sources. Notably, amputee participants achieved similar performance to that of able-bodied participants.

Conclusions: The study therefore shows that the novel feedback strategy conveys useful information about prosthesis movements while reducing the number of motors without compromising performance. This is an important step toward the full integration of such an interface into a prosthesis socket for clinical use.

Keywords: Closed-loop control; Gaussian interpolation; Hannes hand; Multichannel stimulation; Proprioceptive feedback; Spatial encoding; Vibromotors.

Fig. 1

Experimental setup and the scheme of closed-loop control. A: Experimental equipment used; B: Testing session with an amputee; and C: Closed-loop control of Hannes hand and its virtual representation. The subject (3) was seated in front of a monitor wearing the armband with EMG sensors (3b) placed equidistantly around the right forearm, and vibromotors (3a) distributed equidistantly around the interior aspect of the right forearm. The virtual reality interface (2) showed the orientation of the target (2b) and the controlled hand (2c) to implement target achievement test. Importantly, during the test, the subjects controlled a real prosthesis (4). The graphical controls (2a) allowed setting the paramenters of the feedback scheme (5). A PC application (6) governed the training of the pattern recognition model and later controlled the real prosthesis, while the prosthesis state (wrist angle and hand aperture) was conveyed to the subject through vibrotactile feedback. The subject wore headphones (7) to block the incidental noise coming from the Hannes prosthesis (4). More details about the setup and the experimental protocol are provided in the text

Fig. 2

Illustration of the two feedback encoding approaches. A1: Representation of hand aperture range of motion; A2: Representation of the wrist rotation range of motion; B, C and D: Vibromotor activations determined by the two encoding approaches for three different hand configurations (top). The Conventional feedback (middle) used two arrays of vibromotors to separately convey wrist rotation and hand aperture. The Compact feedback (bottom) employs a single array of vibromotors and transmits the feedback information by modulating the Gaussian mean (µ) for wrist orientation and standard deviation (σ) for hand aperture

Fig. 3

Target positions and outcome measures. A: Error between target and reached position (control accuracy) when adjusting the wrist; B and C: Optimal and generated paths and end-point errors for hand and wrist, respectively. Trial time as well as optimal times (t_opt1 and t_opt2) to adjust each DoF are annotated on the x-axes.

Fig. 4

Summary results for the positioning error (A), path (B) and time efficiency (D), and time per DoF (C) in the form of boxplots for able-bodied subjects and two feedback schemes (green – conventional [CN], purple – compact [CM]). HA indicates the aperture and WR indicates the wrist rotation. The small circles are the means, the red lines indicate the medians, boxes are interquartile ranges, whiskers represent min/max values and the crosses are outliers. Dashed vertical lines separate control (left – VF, AF and NF) and test (right – CN and CM) conditions. Horizontal bars show a statistically significant difference between the connected conditions (*, p < 0.05; **, p < 0.01). The asterisks above a condition indicate that the condition was significantly different from all others

Fig. 5

Two example trials performed by an amputee participant using the two feedback methods compared in the manuscript, namely, Conventional (A), and Compact Feedback (B). The plots show the trajectories generated by the participant when adjusting the wrist rotation and hand aperture. Err, PathEff, TimeEff and DoFTime denote positioning error, path and time efficiency, and time per DoF, while the subscripts HA and WR denote hand and wrist, respectively

Fig. 6

The results obtained by 4 amputee participants using Compact and Conventional feedback. Left panels show the performance for each block of trials, to assess the potential learning effect, while the right panels represent the average across blocks. HA indicates aperture, and WR represents wrist rotation. Different colours are associated to different participants (P1-P4). The horizontal dashed lines in the right pannels indicate the mean performance of able-bodied subjects. Bn in the left plots denote different testing blocks