Skip to content
Open access

Quadratic Morphing Morlet Transform and Multi-Scale Frequency-Attentive Temporal Network for sEMG Gesture Recognition

Oct 2026 · Dicle Üniversitesi Mühendislik Fakültesi Mühendislik Dergisi · Vol 17 · 0 citations · 19 references

TL;DR

Experimental results indicate that QMM provides sharper and more discriminative scalograms than classical Morlet, Gaussian, and Mexican Hat wavelets, and a hybrid deep-learning architecture, the multi-scale frequency-attentive temporal network (MFATNet), is proposed as a novel hybrid deep learning architecture.

Abstract

Accurate recognition of surface electromyography (sEMG) signals is essential for intuitive human–machine interaction and myoelectric prosthetic control. However, the inherently transient, nonstationary, and low-amplitude nature of sEMG signals limits the effectiveness of conventional time–frequency representations and deep learning models, which often struggle to capture short-duration neuromuscular events and the multi-scale spectral–temporal characteristics of muscle activity. To address these challenges, this study introduces the quadratic morphing morlet (QMM) wavelet, a newly formulated and fully admissible wavelet designed to enhance transient resolution and spectral selectivity. Building upon this improved representation, a hybrid deep-learning architecture, the multi-scale frequency-attentive temporal network (MFATNet), is proposed as a novel hybrid deep learning architecture. MFATNet integrates tri-branch multi-scale spectral encoding, frequency-wise attention, and dilated temporal modeling to jointly capture spectral diversity, frequency adaptivity, and gesture dynamics within a unified framework. The proposed QMM–MFATNet pipeline is evaluated on the 52-class hand-gesture corpus of the NinaPro DB1 dataset, comprising sEMG recordings from 27 subjects, wherein QMM-based continuous wavelet transform (CWT) representations are used to generate high-resolution scalograms for subsequent model training. Experimental results indicate that QMM provides sharper and more discriminative scalograms than classical Morlet, Gaussian, and Mexican Hat wavelets. Additionally, MFATNet outperforms standard CNN, ResNet, and MobileNet models, achieving the highest accuracy (93.13%) and F1-score (93.15%). Comparisons of different wavelet types further demonstrate the superiority of QMM, which yields a mean accuracy of 0.9313 ± 0.0194, surpassing all traditional kernels.

Read PDF

Similar papers

Open access Oct 2026

EdgeNetTF for lightweight time frequency representation learning in multiclass sEMG hand gesture recognition

Surface electromyography (sEMG)-based hand gesture recognition plays a critical role in human-machine interaction, rehabilitation, and myoelectric control. However, many deep learning models improve recognition performance by increasing architectural complexity, which limits their suitability for wearable and resourc...

Yu-Tao Li, Junghun Kim, Sang-Il Choi · 0 citations
Sep 2026

WP-M2 Net: Wavelet-Perception Macro-Micro Dynamic 1D-CNN for Efficient Sequential Motion Recognition from Sparse sEMG.

Natural and efficient neuromuscular interfaces serve as a vital bridge connecting next-generation neural engineering with wearable human-machine interaction. While sparse surface electromyography (sEMG) is favored for its portability, its limited spatiotemporal resolution poses significant challenges to recognizing hig...

Jun Cheng, Bo Chen, Zhe-Ming Wang et al. · 0 citations
Open access Oct 2026

Lower-limb motion recognition using Gramian angular difference field and multi-scale attention ResNet18 for sEMG signal

Aiming at the problem that traditional methods are insufficient in capturing temporal dependencies for surface electromyography signals, an integrated lower-limb motion recognition framework combining Gramian angular difference field and multi-scale attention ResNet18 is constructed in this paper. First, the Gramian an...

Jiang-Shan Ma, Min-Huan Guo, Yue Zhao et al. · 0 citations
Conference Oct 2026

An action-level gesture recognition method based on sEMG and channel-recalibrated multiscale temporal features

The proposed SE-MSCFormer offers an efficient and robust solution for low-sampling-rate sEMG gesture recognition and confirms that the synergy between the multi-scale temporal configuration and channel recalibration enhances overall recognition performance.

Chao Xian, Lei Liu, Bo Yang et al. · 0 citations
Open access 2026

GestNet: An Adaptive Approach to Multi-Modal Gesture Recognition With Semi-Supervised Learning Through EEG and sEMG Fusion

The ability to reliably decode upper-limb motor intent from biosignals is a key requirement for next-generation Brain-Computer Interfaces (BCIs) and assistive neurorehabilitation systems. In this paper, the authors present GestNet, a semi-supervised deep learning framework for combining surface Electromyography (sEMG)...

Abdollah PourMottaghi, S. A. Zendehbad · 0 citations
Preprint Oct 2026

Interaction-Gated Time-Frequency Learning for Parameter-Efficient PPG-Based Heart Rate Estimation

Photoplethysmography (PPG) enables low-cost and non-invasive heart rate (HR) monitoring but remains sensitive to waveform distortion, physiological variability, and measurement interference. Temporal waveform characteristics and frequency-domain periodicity provide complementary information for HR estimation; however,...

Anh Khoi Pham, Phat Lam, Thien Truong Nguyen Ly et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.