Abstract: The loudness of underwater targets' noise is an important parameter for a sonar operator to classify targets.In the paper,a simple algorithm of specific loudness is proposed.A set of loudness features is extracted,and the classification experiment for three different kinds of targets performed.The results show that the feature extraction method based on the specific loudness of noise signal is useful for underwater target recognition.The difference between loud-ness features and energy features is also analyzed,and it is shown that the convergence of loudness features is better than the latter.