Tactile Myography: An Off-Line Assessment of Able-Bodied Subjects and One Upper-Limb Amputee

Abstract

Human-machine interfaces to control prosthetic devices still suffer from scarce dexterity and low reliability; for this reason, the community of assistive robotics is exploring novel solutions to the problem of myocontrol. In this work, we present experimental results pointing in the direction that one such method, namely Tactile Myography (TMG), can improve the situation. In particular, we use a shape-conformable high-resolution tactile bracelet wrapped around the forearm/residual limb to discriminate several wrist and finger activations performed by able-bodied subjects and a trans-radial amputee. Several combinations of features/classifiers were tested to discriminate among the activations. The balanced accuracy obtained by the best classifier/feature combination was on average 89.15% (able-bodied subjects) and 88.72% (amputated subject); when considering wrist activations only, the results were on average 98.44% for the able-bodied subjects and 98.72% for the amputee. The results obtained from the amputee were comparable to those obtained by the able-bodied subjects. This suggests that TMG is a viable technique for myoprosthetic control, either as a replacement of or as a companion to traditional surface electromyography.

Online Natural Myocontrol of Combined Hand and Wrist Actions Using Tactile Myography and the Biomechanics of Grasping

“…We intentionally decided not to linearise the response in order to save computational effort, and not alter the signal-tonoise ratio. See, e.g., Castellini et al (2018) for more thorough description of the device and its pros and cons.…”

“…The second feature selection method, Gradient-based features extracted from Regions of Interest (ROIs) (Haralick and Shapiro, 1992), has already been used in ultrasound image processing and more specifically to identify finger movements (Castellini et al, 2012;Sierra González and Castellini, 2013;Ortenzi et al, 2015), also together with regression-based algorithms (Castellini et al, 2012;Sierra González and Castellini, 2013). More recently, this method has been further tested in an offline study investigating different methods of feature extraction for the Tactile Bracelet: the ROI gradients gave the highest classification accuracy (Castellini et al, 2018) over Harris corner extraction (Harris et al, 1988) and the structural similarity index (Boschmann and Platzner, 2014) on bicubic interpolated data. Unlike the round-shaped overlapping ROIs used in Sierra González and Castellini (2013) for ultrasound image processing and due to the low resolution of the tactile bracelet compared to ultrasound, a simpler strategy was adopted here after several pre-tests, delimiting each ROI as a non-overlapping 4 × 4 taxel square (Castellini et al, 2018) (a taxel being the value of one sensor), resulting in two ROIs per board (cf.…”

“…More recently, this method has been further tested in an offline study investigating different methods of feature extraction for the Tactile Bracelet: the ROI gradients gave the highest classification accuracy (Castellini et al, 2018) over Harris corner extraction (Harris et al, 1988) and the structural similarity index (Boschmann and Platzner, 2014) on bicubic interpolated data. Unlike the round-shaped overlapping ROIs used in Sierra González and Castellini (2013) for ultrasound image processing and due to the low resolution of the tactile bracelet compared to ultrasound, a simpler strategy was adopted here after several pre-tests, delimiting each ROI as a non-overlapping 4 × 4 taxel square (Castellini et al, 2018) (a taxel being the value of one sensor), resulting in two ROIs per board (cf. Figure 5).…”

“…Such a deformation could be detected by force/pressure sensors and associated to the action, thereby used as an alternative or parallel technique to surface electromyography. Examples of comparisons and mixtures of the two techniques can already be found in literature (Fang et al, 2015;Cho et al, 2016;Connan et al, 2016;Castellini et al, 2018). This approach, involving several independent force sensors has been called, among other ways, force myography or FMG.…”

Special Issue on “Assistive Robotics”

“…They analyzed different Gaussian Process regression kernels on the combination of surface electromyography (sEMG) and tactile myography (TMG) using data from real human experimental subjects, and concluded that the regressed sensor data is effective in proportional control and the detection of prosthetic device activation. In similar research, Castellini et al [6] presented a solution for myocontrol in prosthetic devices using Tactile Myography (TMG). They proposed a tactile bracelet that accommodates different shapes of forearm or residual limb of amputees to measure TMG signals; these measurements are then used to classify differential activation of the wrist or fingers.…”