A Sparsity-Aware Neural Interface with CIM-Based Predictive Focused Sampling for Hotspot Spike Tracking

📋 论文概要

该论文提出了一种基于稀疏感知的神经接口，利用计算存储一体（CIM）引擎进行预测性聚焦采样，实现热点尖峰跟踪。通过结合低功耗全景扫描与高保真尖峰跟踪的动态分配，解决了高密度神经接口的功耗、面积和带宽瓶颈。

💡 主要创新点

• 结合低功耗全景扫描与高保真尖峰跟踪的动态分配机制，实现热点预测与主动位点的高效采样。
• 实时CIM引擎引导的预测聚焦采样，提供16倍分辨率提升（10kHz），显著提高尖峰密度。
• 在40nm CMOS工艺下实现0.0011mm2/ch的高面积效率，降低功耗和硅面积。

📄 原文摘要

Chinese Academy of Sciences, Beijing, China, 4Soochow University, Suzhou, China *Equally Credited Authors (ECAs) 1 3 Abstract This 1024-electrode neural interface solves the power, area, and bandwidth bottleneck by combining low-power panoramic scanning for hotspot prediction with high-fidelity spike tracking on active sites. The dynamic allocation is guided by a real-time compute-in-memory (CIM) engine, providing a 16× resolution boost (10kHz) and 0.0011mm2/ch efficiency in 40nm CMOS. The design maximizes spike density to lower power and silicon area. The pursuit of high-density neural interfaces is fundamentally limited by strict implant power, area, and bandwidth budgets, a constraint exacerbated by the disparity between

👥 作者与机构

Hui Wu*1, Fengshi Tian*2, Jinbo Chen1, Zhipeng Liao1, Hongyong Zhang1, Xing Liu1, Wenjun Zou1, Sirui Cheng1, Zhao Zhang3, Jianlong Xu4, Chi-Ying Tsui2,

Kwang-Ting Tim Cheng2, Jie Yang1, Mohamad Sawan1 Westlake University, Hangzhou, China, 2Hong Kong University of Science and Technology, Hong Kong, China Institute of Semiconductors,