One-Shot Random Forest Model Calibration for Hand Gesture Decoding

Abstract

Objective: Most existing machine learning models for myoelectric control require a large amount of data to learn user-specific characteristics of the electromyographic (EMG) signals, which is burdensome. Our objective is to develop an approach to enable the calibration of a pre-trained model with minimal data from a new myoelectric user. Approach: We trained a random forest model with EMG data from 20 people collected during the performance of multiple hand grips. To adapt the decision rules for a new user, first, the branches of the pre-trained decision trees were pruned using the validation data from the new user. Then new decision trees trained merely with data from the new user were appended to the pruned pre-trained model. Results: Real-time myoelectric experiments with 18 participants over two days demonstrated the improved accuracy of the proposed approach when compared to benchmark user-specific random forest and the linear discriminant analysis models. Furthermore, the random forest model that was calibrated on day one for a new participant yielded significantly higher accuracy on day two, when compared to the benchmark approaches, which reflects the robustness of the proposed approach. Significance: The proposed model calibration procedure is completely source-free, that is, once the base model is pre-trained, no access to the source data from the original 20 people is required. Our work promotes the use of efficient, explainable, and simple models for myoelectric control.

Big Data in Myoelectric Control: Large Multi-User Models Enable Robust Zero-Shot EMG-based Discrete Gesture Recognition

“…Due to the challenges of collecting large datasets, researchers have turned toward few-shot and transfer learning approaches whereby the backbone of the model leverages other users' data, and the end model is fine-tuned to the end user (Jiang et al, 2024;Xu et al, 2024;Côté-Allard et al, 2019). While these approaches can reduce some of the training burden associated with myoelectric control, they are not necessarily ideal for the widespread adoption of EMG.…”

“…Additionally, if not intentionally included in the training procedure, the resulting models lack robustness to confounding factors such as cross-day use, limb-position variation, and electrode shift, thus further exacerbating online usability issues (Campbell et al, 2020). While researchers have been able to alleviate some of these factors through transfer learning (Campbell et al, 2021;Jiang et al, 2024;Xu et al, 2024;Côté-Allard et al, 2019) and domain adaptation (Zhang et al, 2022;Eddy et al, 2023a;Campbell et al, 2024), these strategies fall short from the ideal zero-shot case whereby no training data from the end user is required, such as is the case for current computer-vision based gesture recognition systems (Oudah et al, 2020b).…”