Abstract
Surface electromyography (sEMG) signals carry abundant information regarding human motion and muscle activity, and armbands equipped with sEMG acquisition can decode gestures via pattern recognition algorithms. Consequently, sEMG armbands have been increasingly adopted for building natural and efficient human-machine interfaces. With the expansion of datasets, rapid hardware iteration, and the emergence of multiple sensing modalities, it has become essential to systematically examine how different armband designs and integration strategies affect recognition performance. This paper systematically reviews the architectures and technical specifications of mainstream sEMG armbands and compares the integration and performance of additional modalities-such as inertial measurement unit (IMU), force myography (FMG), magnetomyography (MMG), sonomyography (SMG), near-infrared spectroscopy (NIRS), light myography (LMG), and electrical impedance tomography (EIT)-within sEMG-based systems. This review also highlights the conceptual value of multimodal fusion for improving robustness and generalizability and outlines directions for developing lightweight, low-power, and cost-effective armbands that better support complex human-machine interaction scenarios.