Baseline wander (BW) is a common noise contaminating electrocardiogram (ECG). In order to effectively correct baseline of ECG signal and to preserve more latent components of ECG signal, this paper proposes a simple and novel filter based on statistically weighted moving average. Firstly, after the arrange between the maximum and the minimum of these sampling values in a moving window was divided into many sections with equal width, several segments [a(k), b(k)] including most samples were determined. Then, for every sample point in the moving window its weight was decided according to the criterion: its weight was set as 1 if the sampling value belongs to [a(k), b(k)], otherwise, as 0. Lastly, all these ECG sampling points with 1 weight were averaged to estimate the real baseline in the moving window. The algorithm was tested by simulated signal and real signal from www. physionet. org. The results showed that compared to traditional moving average filter and wavelet package (WP) translation, the proposed filter could more effectively correct baseline in ECG signal and result in less distortion to ECG signal.