Species recognition of sea creatures is very important and is still a difficult task in the assessment of oceanic biological resources by hydroacoustic methods and in optimum selective industrial fishing. Trials have shown that recognition by means of the subjective estimation of echo sounder records and sample hauls is not yet fully satisfactory [1]. In this correspondence a classifier of sea creature species is described. The essential efforts have been made to find efficient procedure of distinguishing features selection. As a result the classifier operates on the basis of an observation vector whose components have been developed in a special way. These components are the central moments of consecutive samples of a few realizations of echo signal envelopes. From the point of view of the observation vector, the classifier is based on simple linear theory. In practice the described classifier can be realized with the aid of the rather uncomplicated microprocessor-based circuits. The chosen distinguishing features concem to exceptional complicated nature of the biological targets. Obtained results indicate that the classifier may also be very useful in the recognition of objects belonging to many nonbiological classes.