A 1.3 μW Event-Driven ANN Core for Cardiac Arrhythmia Classification in Wearable Sensors
Qiao Cai, Xinzi Xu, Yang Zhao, Liang Ying, Yongfu Li, Yong Lian
Abstract
An event-driven system generates samples only when a predefined event is triggered, thus the power consumption tracks the input activities leading to significant savings in power for wearable sensors. In this brief, we presented an event-driven artificial neural network (ANN) core for cardiac arrhythmia classifier (CAC). The proposed data alignment mechanism allows seamless cooperation between the ANN CAC core and the event-driven clockless analog front-end. Measurement results show that the ANN CAC core consumes merely 1.3μW dynamic power at heart rate of 75bpm with a clock frequency of 250kHz a 1.5V supply voltage. Fabricated in 0.18μm, the core occupies 0.75mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> with minimum 1.8 μJ/classification, achieving average classification accuracy of 98% for all 5 types of heart beats as defined in the AAMI.