A novel adaptive median filter, named Adaptive Median Filter based on Fractal Dimension(AMeFFD), is proposed to process analytical signal. A new concept, Relative Point Box Dimension(RPBD), is developed from fractal theory and applied to identify the impulsive noise accurately. According to the results of identification, the window width of median filters (MeF) can be adjusted adaptively. Using the method, the impulsive noise and other common noise can be removed effectively. The filter algorithm has been used to process both simulatied chromatographic signal and practical chromatogram with impulsive noise and other noise. The results showed that, overcoming the drawback of MeF, it was much better than MeF in the RMS index and peak deviation. It has been proved that AMeFFD has the advantage of filtering noise more effectively without distortion of signal peaks and is a powerful tool for processing spectral signal.
