A performance comparison of tonal detectors for low-frequency vocalizations of Antarctic blue whales.

Extraction of tonal signals embedded in background noise is a crucial step before classification and separation of low-frequency sounds of baleen whales. This work reports results of comparing five tonal detectors, namely the instantaneous frequency estimator, YIN estimator, harmonic product spectrum, cost-function-based detector, and ridge detector. Comparisons, based on a low-frequency adaptation of the Silbido scoring feature, employ five metrics, which quantify the effectiveness of these detectors to retrieve tonal signals that have a wide range of signal to noise ratios (SNRs) and the quality of the detection results. Ground-truth data were generated by embedding 20 synthetic Antarctic blue whale (Balaenoptera musculus intermedia) calls in randomly extracted 30-min noise segments from a 79 h-library recorded by an Ocean Bottom Seismometer in the Indian Ocean during 2012-2013. Monte-Carlo simulations were performed using 20 trials per SNR, ranging from 0 dB to 15 dB. Overall, the tonal detection results show the superiority of the cost-function-based and the ridge detectors, over the other detectors, for all SNR values. More particularly, for lower SNRs (⩽3 dB), these two methods outperformed the other three with high recall, low fragmentation, and high coverage scores. For SNRs ⩾7 dB, the five methods performed similarly.

Passive stochastic matched filter for Antarctic blue whale call detection.

As a first step to Antarctic blue whale (ABW) monitoring using passive acoustics, a method based on the stochastic matched filter (SMF) is proposed. Derived from the matched filter (MF), this filter-based denoising method enhances stochastic signals embedded in an additive colored noise by maximizing its output signal to noise ratio (SNR). These assumptions are well adapted to the passive detection of ABW calls where emitted signals are modified by the unknown impulse response of the propagation channel. A filter bank is computed and stored offline based on a priori knowledge of the signal second order statistics and simulated colored sea-noise. Then, the detection relies on online background noise and SNR estimation, realized using time-frequency analysis. The SMF output is cross-correlated with the signal's reference (SMF + MF). Its performances are assessed on an ccean bottom seismometer-recorded ground truth dataset of 845 ABW calls, where the location of the whale is known. This dataset provides great SNR variations in diverse soundscapes. The SMF + MF performances are compared to the commonly used MF and to the Z-detector (a sub-space detector for ABW calls). Mostly, the benefits of the use of the SMF + MF are revealed on low signal to noise observations: in comparison to the MF with identical detection threshold, the false alarm rate drastically decreases while the detection rate stays high. Compared to the Z-detector, it allows the extension of the detection range of ≃ 30 km in presence of ship noise with equivalent false discovery rate.