Self-Powered MXene@Perovskite Thermoelectric Skin for Multimodal Mid-Infrared Sensing and Human Signal Recognition

The effective detection of human thermal radiation (8-14 µm) was critical for next-generation human-machine interactions (HMIs). However, conventional mid-infrared sensors were typically constrained by external power requirements and limited capability in decoding complex multimodal signals. Here, we developed a bio-inspired, self-powered thermoelectric skin based on a V₂C MXene@CsPbBr₃ heterostructure. By combining the local surface plasmon resonance (LSPR) of V₂C MXene with the photothermoelectric effect (PTE) of perovskite, efficient photothermal conversion and self-driven signal generation within the atmospheric window were achieved. The platform supported multimodal signal processing, including gesture recognition, Morse code encryption, and real-time motion tracking (velocity and displacement). In addition, ultrasensitive, bias-free detection of human pulse signals (66-100 bpm) was enabled, and high responsivity, excellent operational stability (>10 000 cycles), and precise capture of subtle physiological dynamics were demonstrated. Furthermore, this skin integrated with a CNN-bidirectional long short-term memory-attention neural network realized silent speech recognition with 95.9% accuracy. This work established a synergistic LSPR-PTE framework, offering a scalable strategy for intelligent electronic skins and contactless HMIs.