Classification of fish schools based on acoustic features associated with tilt angle

Abstract

It is required to monitor fish species and amount in order to save marine resources. In the current study, broadband split-beam echo sounder was used to survey fish resources. The echo sounder was useful to track individual fish in schools because individual fish echoes in the schools could be separately measured and three dimensional positions of the echoes could be accurately estimated. It is well known that the acoustic features of fish echoes are closely related to the body length and the apparent tilt angle of a fish. The tilt angle could be estimated by individual tracking. Echoes from fish schools were measured for several hours in the ocean. Echo signals were gathered into one block within a certain depth and a period of time and analyzed to track individual fish in the schools. Target strength was calculated from the individual fish echo and associated with the tilt angle, which was estimated by the individual tracking result. The feature in each block was statistically calculated by averaging target strengths based on the tilt angles. The blocks of the signals were classified by K-means method with the features and divided into some clusters. As a result of the clustering, fish schools in the blocks could be classified. The distribution of the clusters was dependent on both time and depth.