Review of Multimodal Sensing and Deep Learning in Electronic Skin

电子皮肤是一种柔性电子设备，能模拟人类皮肤的感知功能，其感知能力正从单模态向多模态融合演进，并逐步实现从实验室到商业化的转变。通过集成触觉、化学及温湿度传感器，电子皮肤可感知压力，检测生物标志物，并通过这些功能适应环境变化，在医疗健康监测、智能穿戴和机器人领域展现出巨大潜力。深度学习技术的应用，如卷积神经网络、循环神经网络、Transformers和脉冲神经网络，显著提升了电子皮肤的多模态数据处理效率与智能化水平，推动了其在健康监测和人机交互领域的发展。然而，电子皮肤仍面临多模态信号解耦、生物相容性及能源效率等挑战。未来，随着新型传感器、轻量化算法、仿生材料和自供电技术的引入，电子皮肤有望进一步提高感知精度与经济性。同时，深度学习模型的加入与跨学科合作将加速电子皮肤性能优化与产业化进程。本文综述了电子皮肤的多模态传感技术与深度学习应用，分析了当前挑战，并展望了未来发展方向，为其技术进步与商业化提供参考。Electronic skin (e-skin) is a flexible electronic device capable of mimicking the sensory functions of human skin. Its sensing capabilities are evolving from single-modal to multimodal fusion, gradually transitioning from laboratory research to commercialization. By integrating tactile, chemical, and temperature-humidity sensors, e-skin can perceive pressure, detect biomarkers, and adapt to environmental changes through these functionalities. It demonstrates significant potential in fields such as healthcare monitoring, smart wearables, and robotics. The application of deep learning technologies, such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformers, and Spiking Neural Networks (SNNs), has significantly enhanced the efficiency of multimodal data processing and the intelligence level of e-skin, driving its development in health monitoring and human-machine interaction. However, e-skin still faces challenges such as multimodal signal decoupling, biocompatibility, and energy efficiency. In the future, with the introduction of novel sensors, lightweight algorithms, biomimetic materials, and self-powered technologies, e-skin is expected to further improve sensing accuracy and cost-effectiveness. Additionally, the integration of deep learning models and interdisciplinary collaboration will accelerate the optimization of e-skin performance and its industrialization. This paper reviews the multimodal sensing technologies and deep learning applications of e-skin, analyzes current challenges, and provides insights into future development directions, offering a reference for its technological advancement and commercialization.