A new real-time resource-efficient algorithm for ECG denoising, feature extraction and classification-based wearable sensor network

Abstract

Long-term patient monitoring is an important issue especially for the elderly. This can be done using a wearable wireless sensor network. These sensors have limited resources in terms of computation, storage memory, size and mainly in power. In this work, a real-time resource-efficient algorithm has been implemented and tested practically such that not all the Ephy (ECG) data are transmitted to the server for later processing. The algorithm reads a sample window and processes it on the sensor node using an adaptive filter with a differentiator and then a fast and simple algorithm for feature extraction of the ECG signal to find P, Q, R, S and T waves. Finally, a classifier algorithm has been designed to distinguish between normal and abnormal ECG signals. The work has been implemented using Shimmer sensor nodes and uses the open source TinyOS 2.1.2 and Python 2.7.