Automatic fish sounds classification

Abstract

The work presented in this paper addresses the issue of environmental monitoring. Specifically, it focuses on the use of acoustic systems for passive acoustic monitoring of ocean vitality for fish populations. To this end, various indicators can be used to monitor marine areas such as both the geographical and temporal evolution of fish populations. A discriminative model is built using supervised machine learning (random-forest and support-vector machines). Each acquisition is represented in a feature space, in which the patterns belonging to different semantic classes are as separable as possible. The set of features proposed for describing the acquisitions come from an extensive state of the art in various domains in which classification of acoustic signals is performed, including speech, music, and environmental sounds. Furthermore, this study proposes to extract features from three representations of the data (time, frequency, and cepstral domains). The proposed classification scheme is tested on real fish sounds recorded on several areas, and achieves 96.9% correct classification., compared to 72.5% when using reference state of the art features as descriptors. The classification scheme is also validated on continuous underwater recordings, thereby illustrating that it can be used to both detect and classify fish sounds in operational scenarios.

A Fish and Dolphin Biophony in the Boat Noise-Dominated Soundscape of the Cres-Lošinj Archipelago (Croatia)

“…Sound types needed to be detected, classified and counted, a process that can be achieved manually or automatically [5,39]. If automatic detection tools are better developed for marine mammals than for fish sound types, recent advancement in this sense are encouraging and need to be further expanded [78][79][80]. Here, manual analysis is suggested as the best approach for inspecting and characterizing vocal communities of unknown composition, for which the lack of knowledge on sound type diversity makes automatic detection not feasible.…”

A methodology for analyzing data from long-term passive acoustic monitoring.

“…Nevertheless, it is worth noting that automatic methods are desirable for fish sound detection, especially for long-term acoustic monitoring. Automatic recognition of fish sounds is commonly based on supervised methods (MALFANTE et al, 2016;VIEIRA et al, 2015). As the name implies, supervised algorithms, need to be trained with a set of features obtained from a manually selected training set (JAMES et al, 2013), in this particular case fish sounds.…”

Acoustic classification of Australian frogs for ecosystem survey

“…The sub-band energy of wavelet packet and the bi-spectrum were combined to classify three fish species by the RBF SVM. Malfante et al [2016] presented an automatic fish sounds classification system. Four classes can be detected with an accuracy of 94%.…”