Adaptive ECG Signal Compression Method Based on Look-Ahead Linear Approximation for Ultra Long-Term Operating of Healthcare IoT Devices

Signal compression is an important study in the electrocardiogram (ECG) signal analysis since ECG signal srequire a long time measurement. Linear approximation shows a high signal compression rate, and is efficientin detecting ambiguous fiducial points. Existing research improved the execution time to enable real-time linearapproximation, but the existing algorithm selected the number of vertices arbitrarily. Thus, the existing linearapproximation does not guarantee that the conditions of compression ratio (CR) or reconstruction errormeasured by percentage root-mean-square difference (PRD) will be satisfied. In this study, we improve thealgorithm to enable a linear approximation based on the specified CR or PRD. We propose a quantitativeapproach to determine the optimal number of vertices that satisfies the specified CR through inversecomputation. Additionally, we extend the cost matrix in advance and select the optimal number of vertices ina look-ahead method, thereby performing signal compression according to the PRD. From experimental results,we confirmed an average PRD of 0.78% in the given CR of 10:1, and an average CR of 12.7:1