Abstract
Passive acoustic monitoring has proven to be an indispensable tool for many aspects of baleen whale research. Manual detection of whale calls on these large data sets demands extensive manual labor. Automated whale call detectors offer a more efficient approach and have been developed for many species and call types. However, calls with a large level of variability such as fin whale (Balaenoptera physalus) 40 Hz call and blue whale (B. musculus) D call have been challenging to detect automatically and hence no practical automated detector exists for these two call types. Using a modular approach consisting of faster region-based convolutional neural network followed by a convolutional neural network, we have created automated detectors for 40 Hz calls and D calls. Both detectors were tested on recordings with high- and low density of calls and, when selecting for detections with high classification scores, they were shown to have precision ranging from 54% to 57% with recall ranging from 72% to 78% for 40 Hz and precision ranging from 62% to 64% with recall ranging from 70 to 73% for D calls. As these two call types are produced by both sexes, using them in long-term studies would remove sex-bias in estimates of temporal presence and movement patterns.
Highlights
Calls with a large level of variability such as fin whale (Balaenoptera physalus) 40 Hz call and blue whale (B. musculus) D call have been challenging to detect automatically and no practical automated detector exists for these two call types
The blue and fin whale social call detector and classifier we developed consists of a 2-step modular approach: The first module, a region-based convolutional neural network (rCNN), for detection of ROIs containing potential calls, and the second module, a transformed pretrained Convolutional neural networks (CNN) used to classify the detected ROIs (Fig. 1)
As the number of 40 Hz calls in the DCLDE 2015 was limited, we added additional 1676 forty Hz calls recorded between 7 and 12 June 2008 at yet another site off southern California to the DCLDE data. These data were manually annotated by a trained analyst following a process comparable to that developed for the DCLDE dataset annotation
Summary
Passive acoustic monitoring (PAM), the process of using long-term underwater soundscape recordings to investigate marine life (e.g., Lobel, 2002; Wiggins, 2003; Luczkovich et al, 2008; Tricas and Boyle, 2014; Coquereau et al, 2016) has become an indispensable tool for the study of baleen whale population structure (e.g., McDonald et al, 2006; Balcazar et al, 2015), migration patterns (e.g., Szesciorka et al, 2020) and relative population trends (e.g., Sirovic et al, 2015; Davis et al, 2017) These recordings often cover several months or years, meaning that the detection of calls, if done manually, would be very labor intensive and require significant analyst training effort (Baumgartner and Mussoline, 2011). This paper is a proof of concept for using rCNN in combination with a CNN to detect and classify fin whale 40 Hz and blue whale D calls in PAM recordings
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
