A wearable electrocardiogram (ECG) monitoring device with a customized SoC is reported. The SoC amplifies the ECG signal from passive electrodes and then digitizes and transforms it into wavelet coefficients. A low-power microcontroller (MCU) and a radio frequency (RF) module in the device resolve and send the wavelet coefficients to a mobile platform. The mobile platform uses machine learning algorithms to improve the performance of signal denoising and data compression by exploiting the characteristics of the sensed data, and consequently reduces power consumption in the wearable device. Measurement results show that the device can resolve ECG data from MIT-BIH arrhythmia database and actual ECG signals from human testers. After processing the ECG data with various noise models, the proposed device can improve the signal to noise ratio (SNR) and mean square error (MSE) by 23.8dB and 88.9%, respectively. When resolving actual ECG signals from testers, the typical compression ratio (CR) is 4.8:1 with 1.56% percentage root mean square difference (PRD). The SoC is fabricated in TSMC 0.18 mu m technology, and consumes 45 mu w for different applications.