Assessing Neuromorphic Computing for Fingertip Force Decoding from Electromyography

2025-12-10

AI Summary

This study addresses the low-power requirement for motor intention decoding in neural interfaces by exploring spiking neural networks (SNNs) for high-density surface electromyography (HD-sEMG)-to-fingertip-force regression—a task previously dominated by conventional deep learning models. Method: FastICA is employed to extract motor unit spike trains, and a causal sliding-window training framework is established. The SNN, built upon leaky integrate-and-fire (LIF) neurons and event-driven computation, is benchmarked against an end-to-end temporal convolutional network (TCN). Contribution/Results: The TCN achieves 4.44% MVC RMSE (r = 0.974), while the baseline SNN attains 8.25% MVC RMSE (r = 0.922). Following architectural refinement, the SNN’s accuracy gap narrows substantially, and its superior hardware energy efficiency is demonstrated. This work establishes the first deployable SNN baseline for HD-sEMG force decoding, validating its practical viability and offering a novel neuromorphic paradigm and transferable baseline for next-generation neural interfaces.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

High-density surface electromyography (HD-sEMG) provides a noninvasive neural interface for assistive and rehabilitation control, but mapping neural activity to user motor intent remains challenging. We assess a spiking neural network (SNN) as a neuromorphic architecture against a temporal convolutional network (TCN) for decoding fingertip force from motor-unit (MU) firing derived from HD-sEMG. Data were collected from a single participant (10 trials) with two forearm electrode arrays; MU activity was obtained via FastICA-based decomposition, and models were trained on overlapping windows with end-to-end causal convolutions. On held-out trials, the TCN achieved 4.44% MVC RMSE (Pearson r = 0.974) while the SNN achieved 8.25% MVC (r = 0.922). While the TCN was more accurate, we view the SNN as a realistic neuromorphic baseline that could close much of this gap with modest architectural and hyperparameter refinements.

Problem

• Decode fingertip force from motor-unit firing using HD-sEMG
• Compare spiking neural network with temporal convolutional network
• Evaluate neuromorphic computing for assistive and rehabilitation control

Innovation

• Using spiking neural networks for neuromorphic EMG decoding
• Comparing SNN with temporal convolutional network for force prediction
• Employing FastICA decomposition and causal convolutions for training