Sound-based automatic neotropical sciaenid fishes identification: Cynoscion jamaicensis

Abstract

We present a method to automatically identify sound emissions produced by sciaenid fishes, and in particular by Cynoscion jamaicensis. This species' sound emissions typically have a 24 Hz pulse repetition rate and a quasi-harmonic pattern can be appreciated in spectra, which produces a pitched quality in the sounds. The proposed method is based on long-term partial loudness, pulse repetition rate, pitch strength, and timbre statistics. Current results show an F-measure of 0.9.