Small footprint airborne LIDAR systems now possesses the capability to sample the whole returned waveform rather than to extract discrete 3D coordinate values（discrete point cloud）

thanks to the improvement of data storage hardware and data processing speed.One merit to analyze waveform data is that the end-user can extract point cloud by him/herself from the raw waveform data in the post processing

instead of being provided by the LIDAR system.The first step to analyze waveform data is to decompose the waveform into individual components.Conventional methods for waveform decomposition are usually polynomial fitting by non-linear least square algorithm

or simply thresholding with the threshold value provided by system vendor.Literature has pointed out that it is impossible to get higher accurate decomposition results by such conventional methods.The paper modifies the Expectation Maximum（EM）algorithm in the context of laser scanning waveform decomposition.Experiments with data from both airborne and space borne LIDAR systems show the high reliability and accuracy of the proposed method for waveform decomposition.