Peter J. Dugan

North Atlantic Right Whale acoustic signal processing: Part I. comparison of machine learning recognition algorithms

This paper compares three different approaches currently used in recognizing contact calls made from the North Atlantic Right Whale (NRW), Eubalaena glacialis. We present two new approaches consisting of machine learning algorithms based on artificial neural networks (NET) and the classification and regression tree classifiers (CART), and compare their performance with earlier work that employs multi-Stage feature vector testing (FVT) approach. A combined total of over 100,000 noise and NRW up-call events were used in the study. Calls were primarily recorded from two areas, Cape Cod Bay and Great South Channel. Of the three classifiers, the CART had the highest assignment rates, overall 86.45% with highest false positive rates (≪100 per hour). The FVT Method had exceptionally low false positive rates, with ≪50 per hour. However, it had an overall assignment rate less than the NET. The CART had statistically the same false positive rate as the NET with the highest assignment rates, 2.2% higher than the NET and 11.75% greater than the FVT Method. Details of the results are shown and extensions to the research are discussed.

Seasonal migrations of North Atlantic minke whales: novel insights from large-scale passive acoustic monitoring networks

BackgroundLittle is known about migration patterns and seasonal distribution away from coastal summer feeding habitats of many pelagic baleen whales. Recently, large-scale passive acoustic monitoring networks have become available to explore migration patterns and identify critical habitats of these species. North Atlantic minke whales (Balaenoptera acutorostrata) perform seasonal migrations between high latitude summer feeding and low latitude winter breeding grounds. While the distribution and abundance of the species has been studied across their summer range, data on migration and winter habitat are virtually missing. Acoustic recordings, from 16 different sites from across the North Atlantic, were analyzed to examine the seasonal and geographic variation in minke whale pulse train occurrence, infer information about migration routes and timing, and to identify possible winter habitats.ResultsAcoustic detections show that minke whales leave their winter grounds south of 30° N from March through early April. On their southward migration in autumn, minke whales leave waters north of 40° N from mid-October through early November. In the western North Atlantic spring migrants appear to track the warmer waters of the Gulf Stream along the continental shelf, while whales travel farther offshore in autumn. Abundant detections were found off the southeastern US and the Caribbean during winter. Minke whale pulse trains showed evidence of geographic variation, with longer pulse trains recorded south of 40° N. Very few pulse trains were recorded during summer in any of the datasets.ConclusionThis study highlights the feasibility of using acoustic monitoring networks to explore migration patterns of pelagic marine mammals. Results confirm the presence of minke whales off the southeastern US and the Caribbean during winter months. The absence of pulse train detections during summer suggests either that minke whales switch their vocal behaviour at this time of year, are absent from available recording sites or that variation in signal structure influenced automated detection. Alternatively, if pulse trains are produced in a reproductive context by males, these data may indicate their absence from the selected recording sites. Evidence of geographic variation in pulse train duration suggests different behavioural functions or use of these calls at different latitudes.

North Atlantic right whale acoustic signal processing: Part II. improved decision architecture for auto-detection using multi-classifier combination methodology

Autonomous signal detection of the North Atlantic right whale (NRW), Eubalaena glacialis, is becoming an important factor in monitoring and conservation for this highly endangered species. Both online and offline systems exist to help study and protect animals within this population. In both cases auto-detection of species-specific calls plays a vital role in localizing individual animal by searching time-frequency passive acoustic data. This research presents an experimental system, referred to as the NRW-CRITIC, for automatic detection of the NRW contact call. In general, the CRITIC uses a combinatorial classifier approach to integrate a series of existing machine learning algorithms; each designed specifically for NRW contact call identification. The proposed configuration consists of several recognition methods running in parallel; these include linear discriminant analysis, artificial neural network (NET) and classification regression tree (CART). This paper presents the details for the NRW-CRITIC and discusses the approach used to combine multiple independent decisions into a single result. A side-by-side performance comparison, between the CRITIC and a well-known method, the feature vector testing (FVT), is summarized. Performance metrics are evaluated based on a large database of acoustic recordings consisting of over 58,000 NRW contact calls from various locations, including two critical habitats, Great South Channel and Cape Cod Bay. Results indicate the FVT algorithm yields a 74.7% detection probability with an error rate of 4.35%. In comparison the CRITIC, operating at similar information level yields a 78.02% detection probability with a 3.25% error rate, exceeding the performance of the FVT. Performance was also measured using data from a multi-channel acoustic array located in Massachusetts Bay. A side-by-side comparison of array presence is discussed for two separate days. Results show that with the FVT and CRITIC operating at 0% error for array presence, the FVT method had 18,769 and 24,469 false positives for the Massachusetts Bay datasets respectively. With the same 0% error condition the CRITIC provided successful detection with significantly lower number of false positive rates: 1,072 and 2,324 calls, respectively. Future extensions of this experimental work are also discussed.

Using High Performance Computing to Explore Large Complex Bioacoustic Soundscapes: Case Study for Right Whale Acoustics.

Abstract This paper describes ongoing work to investigate the development of a complex system designed for extracting information from large acoustic datasets. The system, called DeLMA is based on integrating advanced machine learning with high performance computing (HPC). The goal of this work is to provide the capability to accurately detect and classify whale sounds in large acoustics datasets collected using underwater sensors. The case study for this work is focused on detecting the acoustic communication signals of the North Atlantic Right Whale, Eubalaena glacialis , and uses data collected in the Stellwagen Bank National Marine Sanctuary (SBNMS), USA. A summary of the work done for developing a complex detection-classification system and brief description of several algorithms that are used for classifying whale sounds will be covered. A brief discussion on how standard detection algorithms can be incorporated, with no special modifications, into the HPC system for analysis will be mentioned, and two new right whale detection methods are presented, based on continuous region analysis (CRA) and histogram of oriented gradients (HOG). This paper presents a first-hand look at applying the DeLMA system and these algorithms on a large dataset containing over 60,000 channel-hours of acoustic data from the SBNMS. Results from these new detection methods are compared against Baseline algorithms. With the development of the DeLMA system, sound archives can now be explored using a powerful distributed processing architecture. This advancement will allow for rapid execution and visualization of the data using seasonal graphs called diel plots, which show the distribution of detections on a time-of-day vs. time-of-year plane. Diel plots of Baseline, CRA and HOG algorithm results reveal various large-scale features of the seasonality of whale calling behavior. Results are summarized and the authors discuss future areas for study, especially those relate to handling other big passive acoustic data projects.

SEDNA - Bioacoustic analysis toolbox

The possible effects of anthropogenic noise on the marine environment is becoming an important topic in the oceanic community. The exploration for fossil-fuel or alternative energy and the construction of facilities to support these endeavors often requires sizable construction efforts; which usually require permitting to study the impact of noise on the environment. Of particular interest is the variety of data products used to influence environmental impact reports and the processing time required to generate these data from large amounts of passive acoustic recordings. This paper outlines work being done by the Bioacoustics Research Program at Cornell University and the Lab of Ornithology, (BRP) for developing MATLAB tools in support of environmental compliance reporting. Due to the success of acoustic monitoring, understanding acoustic signatures is now becoming part of environmental impact assessment and required compliance for permitting. BRP has leveraged various existing tools and capabilities which result in integrated special purpose software tools within a MATLAB toolbox called SEDNA1. SEDNA incorporates various tools to measure acute and chronic noise levels, detect and classify marine mammal vocalizations, and compute various metrics such as receive levels, signal excess, masking and communication space. This work will summarize the high performance computing strategy used in the SEDNA Toolbox along with the capability to integrate various layers of data within a modeling framework that incorporates ambient noise, vessel and animal data. Finally, the work will demonstrate the power of this approach through animated data visualization, showing animal, vessel and ambient noise integrated over relatively large temporal and spatial scales.

High-Resolution Analysis of Seismic Air Gun Impulses and Their Reverberant Field as Contributors to an Acoustic Environment

In September and October 2011, a seismic survey took place in Baffin Bay, Western Greenland, in close proximity to a marine protected area (MPA). As part of the mitigation effort, five bottom-mounted marine acoustic recording units (MARUs) collected data that were used for the purpose of measuring temporal and spectral features from each impulsive event, providing a high-resolution record of seismic reverberation persistent after the direct impulse. Results were compared with ambient-noise levels as computed after the seismic survey to evidence that as a consequence of a series of repeating seismic impulses, sustained elevated levels create the potential for masking.

Marine acoustic ecologies and acoustic habitats: Concepts, metrics, and realities

Whales, dolphins, and porpoises (cetaceans) are adapted to produce and perceive sounds that collectively span 4–6 orders of magnitude along space, time, and frequency dimensions. Two important concepts, acoustic ecology and acoustic habitat, emerge from this perspective: where acoustic ecology is the study of acoustics involved in interactions of living organisms, and acoustic habitat as ecological space acoustically utilized by particular species. Cetaceans are dependent on access to their normal acoustic habitats for basic life functions. Communication masking from anthropogenic sounds that are chronically present can result in measurable losses of cetacean acoustic habitats, especially for low-frequency specialists, baleen whales. A communication masking model, informed by multi-year datasets, demonstrates cumulative influences of multiple vessels on fin, humpback and right whale acoustic habitats at spatial, temporal, and spectral scales matched to ecologically meaningful habitats. Results quantify ac...