Existing epidermal surface electromyography (sEMG) systems often suffer from a mechanical mismatch at the “soft-hard” interface, where flexible electrodes connect to rigid signal processing units, leading to motion artifacts and interface instability under dynamic deformation. To address this challenge, we present a fully integrated, wireless, stretchable sEMG acquisition system that achieves global mechanical compliance. By employing a multi-layer serpentine interconnect architecture and modulus-matched silicone encapsulation, the system effectively decouples active electronic components from substrate strain, ensuring stable electrical performance under stretching, twisting, and bending. A fractal mesh electrode design with an open-window architecture is utilized to optimize skin-electrode contact impedance. We demonstrate the system’s capability to acquire high-fidelity bioelectric signals in real time across diverse anatomical locations, from large-muscle limb movements to subtle laryngeal vibrations. Subsequent machine learning analysis using a feedforward neural network (FNN) validates the signal quality, achieving a classification accuracy of 99.38% for multi-site limb actions and 93.12% for seven-class laryngeal speech recognition. This work establishes a viable engineering platform for conformable bioelectronics in rehabilitation monitoring and human-machine interaction. • A fully integrated stretchable surface electromyography system is developed using an island-bridge architecture with global mechanical compliance. • Multi-layer serpentine interconnects effectively decouple rigid electronic components from substrate mechanical strain. • Open-window fractal mesh electrodes ensure low-impedance, high-fidelity skin-electrode contact for bioelectric signal acquisition. • The system achieves 99.38% accuracy in limb motion recognition and 93.12% in laryngeal speech recognition via machine learning classification.
Liu et al. (Sun,) studied this question.