Experimental Acoustic Data Analysis for Fish Detection using an Active Ultrasound System

Abstract

This paper proposes the supervised classification of different acoustics signatures employed for fish monitoring by active ultrasound systems. The obtained results are analysed, both qualitatively and quantitatively, by taking into consideration three target types (trout, wood and vegetal debris).

Active Acoustic Signature Extraction by Machine Learning Data Analysis for Underwater Fish Counters

“…Several experimental observations [10], [16], have led us to the machine learning algorithm proposed in this paper for fish recognition. It is composed of two stages:…”

“…The second hypothesis relies on the ability to recognize the external perturbation on the acoustic path [10], [16].…”