ECG Compression Using Subband Thresholding of the Wavelet Coefficients

This paper describes an efficient electrocardiogram (ECG) signal compression technique based on the combination of wavelet transform and thresholding of the wavelet coefficients according to their energy compaction properties in different sub bands to achieve high compression ratio (CR) with low percent root mean square difference (PRD). First, the ECG signal is wavelet transformed using different discrete wavelets. The wavelet transform is based on dyadic scales and decomposes the ECG signals into five detailed band levels and one approximation band level. Then, the wavelet coefficients in each subbands are thresholded using a threshold based on energy packing efficiency (EPE) of the wavelet coefficients. To assess the proper applicability of the proposed technique we have evaluated the effect of threshold levels selection on the quality of the reconstructed signal. To generalize the proposed method, the technique is tested for the compression of a large set of normal and abnormal ECG signals extracted from MIT-BIH database. The performance parameters of the compression algorithm are measured and a CR of 15.12:1 with PRD of 2.5% is achieved. Experiments on selected records from the MIT-BIH arrhythmia database reveal that the proposed method is significantly more efficient in compression than some existing wavelet based ECG compression method. The proposed compression scheme may find applications in digital Holter recording, in ECG signal archiving and in ECG data transmission through communication channels.