Two-Dimensional ECG Signal Compression Based on Region of Interest Using PCA

Several cardiac disorders can be diagnosed by meticulous analysis of ECG signals. The quality of signal determines the accuracy of diagnosis. Usually the size of ECG signals is huge and they are associated with noise. By compressing the ECG signals, they can be stored and transmitted easily. Hence, it is important to pre-process (denoise) and compress it to a maximum extent. In the recent past many works have been done on ECG compression. Compression techniques have been done using time-domain as well transform domain techniques. In this work an approach through Principal Component Analysis (PCA) has been proposed to compress the pre-processed ECG signal and de-compress it efficiently, such that maximum amount of variance is retained. This algorithm has been tested for 28 ECG signals from the MIT-BIH database. In order to analyze the performance of the algorithm, CR (compression ratio) and PRD (Percent Root Mean Square Difference) have been considered as performance parameters and are calculated. The proposed method achieves a good CR along with small PRD in comparison with algorithms that has been proposed by other researchers.