Abstract:
For several years, cardiovascular disease (CVD) has become the major cause of death of non-communicable diseases (NCDs). An electrocardiogram (ECG) is a biomedical signal representing the activity of the heart. ECG monitoring and recording are mainly used for CVD diagnosis. In this research, the lossless ECG compression method is proposed using prediction error-based adaptive linear prediction, which minimizes the ECG signal level to the prediction error value and reduces the code length used for storing digitized data for each sample, and using modified Golomb-Rice coding to encode these prediction error values in binary format and enhance the compression efficiency. In addition, the original signal data is encoded by fixed-length coding. When the signal data is compressed by the proposed method, which is the variable-length coding, the bit length required to store the compressed data varies following the trends of recent samples of the prediction error. For this research, the ECG compression method is tested with the ECG datasets including MIT-BIH Arrhythmia Database (MITDB), PTB Diagnostic ECG Database (PTBDB), and European ST-T Database (EDB). The compression efficiency is indicated by using the compression ratio, which is the ratio of the size of the original signal data to the compressed signal data. The total average compression ratios achieved are 3.533, 3.396, and 3.761 for three datasets respectively. The compression method reduces the space required to store the signal data with no loss or difference from the original signal data. Also, this method efficiently compresses the signal data with different signal acquisition setups including the bit-length determined for each sample, signal resolution, sampling frequency, and duration of acquisition. These features can be implemented for the monitoring and recording of the ECG signal, the diagnostic of CVD, or applied to telemedicine with better performance.