Neuromorphic Hardware Materials for Intelligent Artificial Perception and Multimodal Fusion: From Sensing to Cognition

Biological sensory systems are characterized by the tightly integrated execution of perception, memory, and computation, thereby providing an important model for the development of next‐generation artificial‐intelligence hardware. In contrast, conventional perception systems based on the von Neumann architecture are often limited by intensive data transfer, high‐power consumption, and insufficient real‐time adaptability. To overcome these constraints, neuromorphic devices have been widely explored as a promising hardware platform, because synaptic plasticity and neuron‐like spiking dynamics can be physically emulated, enabling in‐sensor and near‐sensor computing. This review systematically summarizes recent advances in intelligent artificial perception systems based on neuromorphic hardware. The operating principles, material systems, and functional characteristics of two‐terminal memristive devices, three‐terminal neuromorphic transistors, and related integrated architectures are first discussed. Representative progress in neuromorphic visual, tactile, auditory, and chemical perception is then reviewed. Particular attention is also given to multimodal fusion systems, in which heterogeneous sensory information can be temporally coordinated, physically coupled, and jointly processed at the hardware level. Finally, the key challenges that still restrict practical deployment are discussed, including device nonidealities, array‐level uniformity, large‐scale integration, external power dependence, and hardware‐algorithm co‐optimization. Future progress is expected to rely on coordinated advances in materials, devices, circuits, and algorithms. Such efforts will drive neuromorphic artificial perception from single‐function biomimicry toward scalable multimodal cognitive systems.