VLSI implementation for R-wave detection and heartbeat classification of ECG adaptive sampling signals

摘要

Heartbeat classification and R-wave location have been widely used to detect people's health condition. And battery-power devices are mainly applied to wearable scenario, which can monitor people's health all day long, so high energy efficiency is very important under such circumstance. This paper proposes an improved DWT (Discrete Wavelet Transform) algorithm to obtain time-frequency characteristics of adaptive sampling ECG signals to locate the R-waves and compute RR-intervals, and further applies KNN (K-Nearest Neighbor) as a classifier to classify heartbeats, then presents a power-efficient VLSI architecture for the proposed algorithms. Experiments with the MIT-BIH Database show that the proposed DWT algorithm obtains Se (sensitivity) of 96.29% and Pp (positive predictive) of 90.26% compared to 66.44% and 70.43% respectively by the original DWT algorithm under noise level of SNR=8.And the accuracy of the heartbeat classification is 87.7%. The VLSI is implemented using SMIC 65nm CMOS technology and the power consumption is 112.7μ W at 2 alternate frequencies of 60 and 360 Hz.