ECG monitoring has high requirement on reconstruction accuracy for ECG signal, and is an important application of WBAN where existing problem of low power consumption. Although power consumption is reduced by current ECG reconstruction algorithm, but frequency domain characteristics of ECG signal are not fully utilized. In this paper, a method of ECG denoising and reconstruction for wireless body area network using compressed sensing based on thresholding is proposed. Based on CS theory, ECG signal is sampled in sensor node using a fixed measurement matrix, then it is denoised and reconstructed in sink node using a fast reconstruction algorithm based on thresholding after transmitted to sink node. Simulation result shows that this method has advantages such as high reconstruction accuracy, fast reconstruction and high performance of denoising.