The Airborne laser scanning (ALS) system has been one of the most effective and reliable means of bare earth collection in the recent decade. The better penetration characteristic of LiDAR system compared with Photogrammetry technique makes the ALS system more practical for DEM generation in a forest area. By filtering the non-ground points out of the LiDAR data, the DEM can be simply obtained by the remaining ground points. Normally the multi-return system can provide up to 4 points each laser pulse. However the information of how the points are exactly extracted is usually not provided by the LiDAR system. Thanks to the development of LiDAR technique, the commercial ALS systems now can record rich return laser signals called waveforms. A waveform contains the information of the interaction between targets and the emitted pulse along the laser lighting path. This provides users to develop their approaches to extract points from waveform data. Normally the number of the points extracted from waveform data is more than the number of points provided by the multi-return system because some weak echoes or overlapping echoes can be further detected from waveform data. For the use of DEM generation, those increased points can be the key features for the description of terrain relief. This paper utilizes a wavelet-based approach to detect the points from waveform data. We compared the generated DEMs from the two kinds of data sources: multi-return points and waveform points. Our preliminary results show the wavelet-based detector can detect the weak or overlapping echoes effectively. The number of points is 20% more than the number of points provided by the multi-return system. As the ground points increase, the accuracy of DEM also increases especially in some steep areas.