Discrete Hilbert Transform via Memristor Crossbars for Compact Biosignal Processing
Discrete Hilbert Transform via Memristor Crossbars for Compact Biosignal Processing
复制标题
DOI：
10.1109/aict55583.2022.10013604
复制
发表时间：
2022-10
期刊：
影响因子：
--
通讯作者：
Lei Zhang;Zhuolin Yang;Kedar K. Aras;Igor R. Efimov;G. Adam
中科院分区：
文献类型：
--
作者：
Lei Zhang;Zhuolin Yang;Kedar K. Aras;Igor R. Efimov;G. Adam
希尔伯特变换在生物医学信号处理中有着广泛的应用，需要高效的实现。提出了基于新兴忆阻器器件的离散希尔伯特变换的实现方法。它使用在忆阻器阵列中编程的权重的两个矩阵乘法层和可映射到CMOS的线性Hadamard乘积计算层。该功能在来自人类心脏的光学心脏信号的数据集上进行了测试。结果表明，所提出的实现方法与MATLAB函数之间的夹角误差可以忽略不计。它对非理想情况也具有稳健性。该方案可应用于边缘生物信号处理。
The Hilbert transform is widely used in biomedical signal processing and requires efficient implementation. We propose the implementation of the discrete Hilbert transform based on emerging memristor devices. It uses two matrix multiplication layers using weights programmed in the memristor array and a linear Hadamard product calculation layer mappable to CMOS. The functionality was tested on a dataset of optical cardiac signals from the human heart. The results show negligible <1% angle error between the proposed implementation and the MATLAB function. It also has robustness to non-idealities. This proposed solution can be applied to bio-signal processing at the edge.