ECG Classification Using Knn And LDA for Continuous Heart Monitoring

Abstract

In this work, an ECG data representation and encoding schema is investigated. Its aim is to support mobile and continuous heart monitoring, for athletes and cardio-vascular disease (CVD) patients.

For data analysis and encoding, a linear discriminant analysis (LDA) was performed on ECG data capturing several heart conditions, obtained from PhysioNet. On-line performance, in terms of classification of unknown heartbeats, using k-nearest neighbours (kNN), was computed and reported.

We show that such an approach allows for simple, well-established, and robust data classification tools to be deployed. Using this representation schema, we hypothesize there is a potential for cheaper and more user- friendly apparatuses in the market.