The question of the ECG data compression is Olle of tile inost important subject ill biosignal processing. If BI' Artificial Neural Network is adopted to compress ECG data, the compression precision and the compression ratio are high, but the compression speed is low. The system we estabiished improves the speed by taking some nleasures. The system adopts four filters to process ECG data for improving the quality of ECG signals, and adopts improved first dilllerence algorithm to detect R points accurately and easily. These works benefit the following compression. During the compression, we combinc BP algorithm with TP algorithm and establish a weight template library to improve the rate of compression. The compression precision and compression ratio Call meet the reqtliremellt and the processing of ECG data containing a few types of abnormal ECG waves is real time with the system.