In China, up to now, the homemade depth bomb weapons are equipped with traditional contact fuze, that has low effectiveness in attacking the enemy's submarine, and can hardly meet the requirement of modern naval warfare. In overseas develop countries; however, the service depth bomb weapons are equipped with non contact fuze that has high effectiveness in attacking underwater craft. There is remarkable gap between China and overseas develop countries in depth bomb fuze technology. So, at present, strong emphasis is laid on the development of depth bomb proximity fuze in China. Our national Eleventh-five mutual application techniques advanced research project in fuze and Initiating Explosive Device is developing the research of depth bomb ultrasonic proximity fuze.Target feature extraction plays an important role in the depth bomb fuze system, because it is a key point that ensures the fuze system of successfully recognizing target. In the process of target detection, depth bomb fuze will undergo the interference from echo signal of neighborhood bomb that is the primary factor which causes high false-alarm probability. This thesis aims at studying the question of neighborhood bomb interference, and giving a solving method, which is expected to extract the feature of echo signal, distinguish the target echo signal from the neighborhood bomb echo signal, and eventually eliminate the neighborhood bomb interference. The main contents are as follows:1. Underwater target feature extraction methods used in the active sonar system are inducted and researched. The characteristic of the echo wave signal, received by the depth fuze, is analyzed. Then an target feature extraction method based on the Doppler shift principle. And the power spectrum estimation is chosen to implement the target feature extraction method.2. Several typical power spectrum estimation methods that are used at present is inducted in detail.3. Through computer emulation, all the methods that are inducted before are analyzed in compare with each other. And the validity of the chosen methods is tested in a working condition that is constructed in computer. And all the concerned index parameter is fixed.