This paper presented an ECG automatic detection system, including analog front-end acquisition circuit and digital recognition algorithm, which is used to collect and handle the ECG signal of human body. According to the analysis results, the real time health status of human body can be obtained directly. The low frequency baseline drift is suppressed by the digital-analog mixing method, based on Savitzky-Golay (SG) Filter, and the characteristic morphological of ECG signals are analyzed in real time by feature recognition algorithm, based on adaptive threshold. The system detects QRS wave through MIT-BIH arrhythmia database and T wave through QT database. The accuracy of QRS and T wave detection is over 99% and 94% respectively.