Highly Defined Whale Group Tracking by Passive Acoustic Stochastic Matched Filter

Abstract

In this paper, we compare two low cost time-domain tracking algorithms based on passive acoustics. The problem consists in tracking an unknown number of sperm whales (Physeter catodon). Clicks are recorded on two datasets of 20 and 25 minutes on an open-ocean widely-spaced bottom-mounted hydrophone array. The output of the method is the track(s) of the Marine Mammal(s) (MM) in 3D space and time. Firstly, we briefly review studies of the Stochastic Matched Filter (SMF) detector and its performances with a reflected click cancellation, the Teager-Kaiser-Mallat (TKM) filtering, the source separation methods and the main characteristics of MM signals. Then, we propose a real-time algorithm for MM transient call localization. We also recall the Cramér-Rao Lower Bound (CRLB) Kay (1993) and the confidence ellipses theory to predict the reachable accuracy and compare it to the tracking results. In Section 3 we show and compare results of track estimates with results from specialized teams and compare SMF versus TKM localization. Then, the system is evaluated with the confidence ellipses on the trajectories. Finally, we discuss on the possible dynamic behavior of the whale that these localizations offer, like hunting and foraging strategies. This paper deals with the 3D tracking of MM using a widely-spaced bottom-mounted hydrophone array in deep water. It focuses on sperm whale clicks. There were previous algorithms developed in the state of the art Giraudet & Glotin (

Passive stochastic matched filter for Antarctic blue whale call detection

“…For the MF, the observation is previously filtered by the application of a band-pass filter in the Z-call frequency band (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). The detection threshold is set at Ts ¼ 0.01.…”

“…where f c ¼ 22.6 Hz is the central frequency in the [15][16][17][18][19][20][21][22][23][24][25][26][27][28][29][30] Hz bandwidth, L and U are, respectively, linked to the lower and upper asymptotes of the Z-call, M represents the time shift and a the grow rate. The amplitude a(n), is set to vary in accordance with the energetic difference between unit A and C (Sec.…”

“…The SMF has been used for the passive acoustic detection of whales clicks. 15,28 In those studies, the noise covariance matrix is estimated on annotated signal free samples. Although it is quite convenient for supervised detection on relatively short records, it becomes tedious and unpractical for automatic detection on several-hours-long passive acoustic monitoring datasets with highly varying background noise.…”

“…It relies on an extensive computation over an artificial dataset, and has been realized for the classic SMF in several publications. 15,28,31 Yet, performance simulations might not be representative of the method's robustness to noise in real conditions. The third option is to confront the algorithm to a ground truth dataset, 17,35,37 keeping in mind that the annotating process has been shown to be subjective and quite variable, especially in low SNR conditions.…”

“…The SMF was originally derived from the MF for active sonar processing to maximize the SNR of stochastic signals embedded in colored noise. These assumptions are also well suited for passive detection of stereotyped marine mammals calls 15,16 where emitted signals are modified by the unknown impulse response of the propagation channel. Therefore, at the reception, they can be considered as stochastic.…”