Multichannel Wearable Sensor Based on Porous Structure for Simultaneous Acquisition of Different Mechanical Signals

The intricate mechanical signals inherent in human motion pose significant challenges for the accurate and real-time recognition of postures, necessitating the development of sensors capable of detecting multiple mechanical stimuli concurrently. Herein, a resistance and capacitance multichannel sensor with porous structure is reported. Benefitting from the pores introduced by the prefoaming and steeping process, the resistance channel exhibits heightened sensitivity to strain (gauge factor, GF, of 2.112 for 0–80% strain and 14.227 for 80–150% strain) while remaining nearly insensitive to pressure. Conversely, the capacitance channel demonstrates high sensitivity to pressure (sensitivity, S, of 41.46 Pa–1 for 0–5.5 kPa pressure) and significantly lower sensitivity to strain. The sensor’s superior anti-cross-talk capability allows for the precise differentiation of motion characteristics augmented by machine learning algorithms, enabling the accurate identification of various joint and finger types and states with high recognition rates of 99.89% and 98.56%, respectively. Additionally, a Tai Chi posture recognition system has been developed, leveraging a lightweight hybrid convolutional neural network–long-term memory (CNN-LSTM) model, achieving a remarkable classification accuracy of nearly 99.85% for four distinct continuous Kungfu forms.