A 65nm Neuromorphic Bio-Signal Encoder with Compute-in-Entropy Architecture 7.13nJ Privacy-Preserving Encoding and 2.38Mb/mm2 Item Memory Density

This paper presents a neuromorphic bio-signal encoder for privacy-critical smart health applications featuring a novel compute-in-entropy architecture for hardware-efficient, physically-unclonable hyperdimensional encoding. The test chip achieved privacy-preserving bio-signal encoding with 7.13 nl efficeincy, 2.38 Mb
$/\text{mm}^{2}$
item memory density, and in-situ decision-making and learning capabilities.