On marine mammal acoustic detection performance bounds

Abstract

Abstract-Since the spectrogram does not preserve phase information contained in the original data, any algorithm based on the spectrogram is not likely to be optimum for detection. In this paper, we present the Short Time Fourier Transform detector to detect marine mammals in the time-frequency plane. The detector uses phase information for detection. We evaluate this detector by comparing it to the existing spectrogram based detectors for different SNRs and various environments including a known ocean, uncertain ocean, and mean ocean. The results show that this detector outperforms the spectrogram based detector. Simulations are presented using the polynomial phase signal model of the North Atlantic Right Whale (NARW), along with the bellhop ray tracing model.

Whistle detection and classification for whales based on convolutional neural networks

“…Typically, these methods follow the following steps: sound preprocessing, whale sound detection, feature extraction of detected sounds, and feature classification. Among them, Short Time Fourier Transform (STFT) [11][12][13][14][15], Wavelet Transform (WT) [16] and Hilbert Huang Transform (HHT) [17] were used to extract features of whale sounds. Artificial Neural Network (ANN) [13,16], Support Vector Machine (SVM) [11,17] and Sparse Representation-based Classifier (SRC) [18] were used for classifying the extracted features.…”