Visual Feedback of Pattern Separability Improves Myoelectric Decoding Performance of Upper Limb Prostheses

TL;DR

The Reviewer is introduced, a 3D visual interface projecting EMG signals directly into the decoder's classification space, providing intuitive, real-time insight into PR algorithm behavior, facilitating immediate and consistent PR-based myoelectric control improvements.

Abstract

State-of-the-art upper limb myoelectric prostheses often use pattern recognition (PR) control systems that translate electromyography (EMG) signals into desired movements. As prosthesis movement complexity increases, users often struggle to produce sufficiently distinct EMG patterns for reliable classification. Existing training typically involves heuristic, trial-and-error user adjustments to static decoder boundaries. Goal: We introduce the Reviewer, a 3D visual interface projecting EMG signals directly into the decoder's classification space, providing intuitive, real-time insight into PR algorithm behavior. This structured feedback reduces cognitive load and fosters mutual, data-driven adaptation between user-generated EMG patterns and decoder boundaries. Methods: A 10-session study with 12 able-bodied participants compared PR performance after motor-based training and updating using the Reviewer versus conventional virtual arm visualization. Performance was assessed using a Fitts law task that involved the aperture of the cursor and the control of orientation. Results: Participants trained with the Reviewer achieved higher completion rates, reduced overshoot, and improved path efficiency and throughput compared to the standard visualization group. Significance: The Reviewer introduces decoder-informed motor training, facilitating immediate and consistent PR-based myoelectric control improvements. By iteratively refining control through real-time feedback, this approach reduces reliance on trial-and-error recalibration, enabling a more adaptive, self-correcting training framework. Conclusion: The 3D visual feedback significantly improves PR control in novice operators through structured training, enabling feedback-driven adaptation and reducing reliance on extensive heuristic adjustments.

Figures (7)

• Figure 1: Motor-based training for PR-based prostheses. The process is structured into two parts: calibration and exploration. In calibration, users are prompted to perform muscle contractions as EMG signals are recorded. Features are extracted from these EMG signals and are used to train a PR-based motion classifier. During exploration, users observe classification output through modalities like a virtual arm and assess control performance. During this period, users can improve the quality of their control by: 1) adjusting muscle activity based on feedback; or 2) returning to the calibration phase to re-initialize the motion classifier with newly recorded EMG.
• Figure 2: Representation of visualization feedback when example gestures are performed. (A-C) The Reviewer: Using the PR-based control algorithm, EMG patterns for different gestures are clustered based on features extracted during the calibration phase. The resting position is marked as the initial point (0, 0, 0) in the feature space. Guided by patient data, the system optimizes the distribution of entries to maximize interclass variance and minimize within-class variance. After calibration, when the user performs a gesture, the white cursor moves to the corresponding location in the feature space as classified by the PR system. Subjects in the Experimental Group can access this system during the exploration phase and rotate the space for a comprehensive view. (D-F) Virtual Arm: When a gesture is performed, this system shows real-time movements of the arm as classified by the PR system. The Control Group accesses the virtual arm during the exploration phase.
• Figure 3: Experimental setup. (A) Equipment: i) A bebionic small multi-articulated hand was attached to each subject to simulate the weight-bearing experience of individuals with ULL. ii) A numbered board guided user movements, while iii) a display screen on the user's right showed gesture icons during calibration to prompt corresponding movements. In the exploration phase, the Experimental Group used a 3D visual system, while the Control Group used a virtual arm. (B) The FLT was administered during the testing phase, with subjects instructed to keep their dominant limb's shoulder aligned with Location 5, as indicated by the board. During testing, subjects placed the multi-articulated hand of the prosthesis in one of three testing locations. Left-handed individuals rotated between Locations 3, 4, and 5, while right-handed individuals alternated between Locations 1, 5, and 6.
• Figure 4: Overview of the Fitts Law Test. (A) Subjects were tasked with aligning a virtual cursor to a target aperture and orientation, displayed as a red ring and protrusion. A trial was deemed successful only when both the ring and protrusion were aligned accurately within a timeframe of 15 seconds. (B) In the FLT, subjects would maneuver a black ring along with a protrusion on the ring. The gestures designated for commanding the ring and protrusion are illustrated above. Upon commencement of a trial, users would be prompted with a gesture to control the ring's closure. The prompted gestures encompass Power Grasp, Key Grasp, Tripod Grasp, Index Point, and Precision Pinch.
• Figure 5: Results of the Experimental and Control Group across the ten-session longitudinal study. In Sessions 1 -- 4, participants were asked to calibrate four active movements (and Rest). In Session 5, two additional movements were incorporated. In Session 8, two further movements were added to increase task difficulty. Throughout each session, the following metrics were computed for each participant: (A) Completion Rate; (B) Overshoot; (C) Path Efficiency; (D) Throughput; (E) Normalized Training Time; and (F) Number of Recalibrations. Asterisks indicate sessions wherein the difference between the Experimental Group and Control Group was significant ($p < 0.05$; Student's t-test).