Application of Context-dependent Interpretation of Biosignals Recognition to Control a Bionic Multifunctional Hand Prosthesis

TL;DR

This paper addresses limited dexterity in sEMG-driven hand prostheses by introducing a context-dependent recognition framework built as a multiclassifier system of isolated boxes. It formalizes encoder/decoder relationships, defines box structures, and proposes two optimization problems ($Q_l$ per box; global $Q$ via secondary-meaning permutation) solved by exhaustive search or evolutionary algorithms. Experimental results on multiple sEMG datasets show context-dependent methods achieve similar or better classification quality than context-free baselines while substantially increasing the prosthesis's movement repertoire. The approach is non-invasive, scalable, and offers a practical route to enhance dexterity without relying on invasive surgical interventions.

Abstract

The paper presents an original method for controlling a surface-electromyography-driven (sEMG) prosthesis. A context-dependent recognition system is proposed in which the same class of sEMG signals may have a different interpretation, depending on the context. This allowed the repertoire of performed movements to be increased. The proposed structure of the context-dependent recognition system includes unambiguously defined decision sequences covering the overall action of the prosthesis, i.e. the so-called boxes. Because the boxes are mutually isolated environments, each box has its own interpretation of the recognition result, as well as a separate local-recognition-task-focused classifier. Due to the freedom to assign contextual meanings to classes of biosignals, the construction procedure of the classifier can be optimised in terms of the local classification quality in a given box or the classification quality of the entire system. In the paper, two optimisation problems are formulated, differing in the adopted constraints on optimisation variables, with the methods of solving the problems based on an exhaustive search and an evolutionary algorithm, being developed. Experimental studies were conducted using signals from 1 able-bodied person with simulation of amputation and 10 volunteers with transradial amputations. The study compared the classical recognition system and the context-dependent system for various classifier models. An unusual testing strategy was adopted in the research, taking into account the specificity of the considered recognition task, with two original quality measures resulting from this scheme then being applied. The results obtained confirm the hypothesis that the application of the context-dependent classifier led to an improvement in classification quality.

Figures (10)

• Figure 1: Open-loop control system for the bioprosthesis of the upper limb
• Figure 2: An example of the structure of the context-dependent classification system
• Figure 3: Time scheme of human and prosthesis activity inside a box (for the sake of simplicity, all classification results are referred to their interpretation, i.e. to the intention of the movement). Events: (1) - onset detection; (2) - offset detection or timeout $T$; (3) - recognition of the intent to perform the box-opening movement; (4) - performing recognised box-opening movement; (5) - recognition of the intent to perform a movement inside the box; (6) - performing a recognised movement inside the box; (7) - recognition of the intent to perform the box-closing movement; (9) - performing a recognised box-closing movement; bringing the prosthesis to its initial state. Prosthesis activities: (A) - prosthesis in the initial position (state) - e.g. the resting position; (B) - recording of detected sEMG biosignals; (C) - operation of the classifier external to the box; (D) - operation of the kinematic control algorithm; (E) - operation of the classifier inside the box; (O) - the prosthesis is listening. Human activities: (K) - generating sEMG biosignals indicating the intent to perform the box initiating/closing movement; (Li) - generating sEMG biosignals indicating the intent to perform subsequent movements inside the box.
• Figure 4: Example of the structure of the context-dependent classification system. Types of movements Cini2019: (m1) pronation; (m2) supination; (m3) oblique grip; (m4) hook grip; (m5) spherical grip; (m6) cylindrical grip; (m7) precision grip; (m8) key grip; (m9) wrist flexion; (m10) wrist extension; (m11) index finger flexion; (m12) ring index flexion; (m13) finger point; (m14) mouse grip; (m15) lateral grip; (m16) platform grip. The symbol $(+)$/$(-)$ means that the movement initiates/closes the box. Boxes are marked with a double line.
• Figure 5: Example 1: (A) structure of the context-dependent classification system (box arrangement); (B) permitted and non-permitted class numbers for the secondary interpretation; (C) the solution tree.

Theorems & Definitions (2)

• Example 1
• Example 2