Study On Ambulatory Electrocardiograph Data Compression Method

The application and popularization of ambulatory electrocardiograph (Ambulator ECG) is an effective approach to improve the prevention and treatment of cardiovascular diseases. The characteristics of Ambulator ECG, such as long-time record and multileads, lead to the large volume of data, and create difficulties for storage and remote transmission. This promotes the requirement of data compression.First of all, this paper denoised the original Ambulator ECG by introducing a four-point averaging filter and putting forward improving for some signal defects. The high-frequency noise was removed and the baseline drift was corrected.By taking advantage of the wavelets in signal singularity detection, this paper detected R waves in Lead V5. Combined with other detection methods, the undetected, over-detected or false detected R waves were corrected and the robustness of this detection algorithm has been improved. The different data leads among Ambulator ECG were synchronously recorded, so the result from the R waves detection algorithm also applied to other leads. Feature waves and non-feature waves were separated from each other in a complete data lead based on the result from the R waves detection.Because there is much abundant important diagnostic value contained in QRS complexes, it is regarded as feature waves, and losslessly compressed by Huffman coding. By choosing appropriate downsampling factor, the downsampling of non-feature waveform effectively reduced the amount of data. Also, application to the downsampling result with cubic interpolation algorithm for reconstruction could restore P wave and T wave and other diagnostic information in non-feature waveform with high fidelity. To further improve the compression efficiency, Huffman coding was applied to the downsampling result.Einthoven Law was introduced to remove the redundancy among 12 data leads, and the reliability of this algorithm was scientifically validated.Besides, some algorithm evaluation indexes were introduced in this paper. Study found that the compression efficiency and the reconstruction error were mainly affected by the downsampling factor. The error and CR (compression ratio) were calculated, and the change of error and CR with the downsampling factor was described. The result shows that: when downsampling factor is not greater than 5, the reconstructed P-QRS-T waveform is of better diagnostic information, and the global CR achieves between 11.5 and 17.8, which means the compression performance is superior to other similar algorithms.The computational complexity was analyzed at the end of this paper. The computational complexity of the major parts of this algorithm is O ( N ), which means the fastest execution speed.