油藏计算
计算机科学
神经形态工程学
可穿戴计算机
特征(语言学)
实时计算
心律失常
事件(粒子物理)
人工智能
频道(广播)
数字传感器
心脏监护
功率消耗
财产(哲学)
信号处理
计算
记忆电阻器
数据流挖掘
特征提取
深度学习
人工神经网络
采样(信号处理)
边缘计算
功率(物理)
可穿戴技术
模式识别(心理学)
大数据
块(置换群论)
作者
Hanrui Li,Dayanand Kumar,Nazek El‐Atab
标识
DOI:10.1002/aisy.202500350
摘要
Cardiac arrhythmia detection is crucial for preventing heart‐related emergencies with wearable electrocardiograph (ECG) monitoring and recording. However, traditional systems face significant challenges in terms of power consumption and real‐time data processing. This article presents a novel memristor‐assisted event‐driven system for low‐power and efficient ECG signal processing. The system incorporates a memristor‐based level‐crossing event encoder, reducing redundant data by up to 88.23% compared to traditional Nyquist sampling methods. These event streams are then processed by a cascaded memristor‐based reservoir layer, which dynamically maps temporal patterns into a high‐dimensional feature space, enabling robust and efficient feature extraction. Additionally, a memristor crossbar array is employed to accelerate matrix computations for neuromorphic network classification. Leveraging the inherent low‐power and non‐linear dynamics of memristor devices, the proposed system achieves high arrhythmia detection accuracy (up to 98.43%) with an average power consumption of 5.26 mW, demonstrating its potential for real‐time, wearable edge AI cardiac monitoring applications.
科研通智能强力驱动
Strongly Powered by AbleSci AI