Stochastic Sequential Sensory Selection for Gesture Recognition in KineticoMyoGraphy Guided Bionic Hands
Artikel i vetenskaplig tidskrift, 2025

KineticoMyoGraphy (KMG) is an emerging sensor technology offering innovative solutions for tracking amputees' fine muscle movements, promising better hand gesture recognition with greater sustainability than existing methods. The primary challenge in KMG technology lies in the required number and placement of magnetic sensors to balance accuracy, sustainability, and cost-efficiency for practical hand gesture interpretation. To tackle this issue, we propose a Stochastic Sequential Strategy for Magnetic Sensory Selection (S3MSS). We apply this strategy to a configuration of 16 magnetic sensors surrounding surgically implanted magnets in a patient's forearm. The method uses an Error-Correcting Output Codes (ECOC) framework with Multiclass Linear Discriminant Analysis (MCLDA) and Multiclass Support Vector Machines (MCSVM). Our approach emphasizes robust sensory selection and consistent performance through time-based seeding and K-fold cross-validation. Clinical results indicate consistency in sensory selection across two independent trials, underlining this factor as crucial for reliability. Statistical significance test confirms the superiority of the MCLDA over the MCSVM approach, achieving a 93% accuracy in the classification of Fingers, Wrist, and Thumb gestures using only five sensors near the magnets' motion range. This underscores our strategy's effectiveness in accurately detecting hand movements, highlighting its potential for clinical application and improving amputees' quality of life.

Medical robotics

Electromyography

KMG signals

Bionic hands

Muscles

Robot sensing systems

Electroencephalography

Biomimetics

gesture recognition

Accuracy

Motors

Magnets

sequential selection

magnetic sensors

gesture recognition

Författare

Arman Abasian

Ferdowsi University of Mashhad

Hamed Farhadi

Chalmers, Elektroteknik, System- och reglerteknik

Ferdowsi University of Mashhad

Mohammad-R. Akbarzadeh-T.

Ferdowsi University of Mashhad

Alireza Akbarzadeh

Ferdowsi University of Mashhad

Ali Moradi

Mashhad University of Medical Sciences

Amir-M. Naddaf-Sh

Lamar University College of Engineering

IEEE Transactions on Medical Robotics and Bionics

25763202 (eISSN)

Vol. 7 1 325-336

Ämneskategorier (SSIF 2025)

Elektroteknik och elektronik

DOI

10.1109/TMRB.2024.3503993

Mer information

Senast uppdaterat

2025-03-28