Abstract
In IoT applications, it is often necessary to achieve an optimal trade-off between data compression and data quality. This study investigates the effect of Compressed Sensing and reconstruction algorithms on ECG arrhythmia detection using SVM classifiers. To neutralise the mutual effect of compression and reconstruction algorithms on one another, we consider each reconstruction algorithms with various compression ratios and vice versa. The employed reconstruction algorithms are Basis Pursuit (BP) and Orthogonal Matching Pursuit (OMP). We employ two steps: (a) identifying proper compression ratio that withholds essential information of ECG signals, (b) assessing the impact of two reconstruction algorithms and their exactness on quality of classification. The findings of this study are threefold: (a) Remarkably, the SVM classifier requires few samples to detect ECG arrhythmia. (b) The results indicate for compression ratios up to around 1:7 ECG signals are recovered then classified with the same quality for both algorithms. However, by increasing compression ratio BP outperforms OMP in terms of ECG arrhythmia detection. (c) Negative correlation between compression ratio and signal quality is observed, that is intuitive enough to realise the trade-off between them.