Tyler Helble

Followers

Following

Co-authors

Public Views

Interests

Uploads

Papers by Tyler Helble

Ecological inferences about marine mammals from passive acoustic data

Biological Reviews

ABSTRACTMonitoring on the basis of sound recordings, or passive acoustic monitoring, can compleme... more ABSTRACTMonitoring on the basis of sound recordings, or passive acoustic monitoring, can complement or serve as an alternative to real‐time visual or aural monitoring of marine mammals and other animals by human observers. Passive acoustic data can support the estimation of common, individual‐level ecological metrics, such as presence, detection‐weighted occupancy, abundance and density, population viability and structure, and behaviour. Passive acoustic data also can support estimation of some community‐level metrics, such as species richness and composition. The feasibility of estimation and certainty of estimates is highly context dependent, and understanding the factors that affect the reliability of measurements is useful for those considering whether to use passive acoustic data. Here, we review basic concepts and methods of passive acoustic sampling in marine systems that often are applicable to marine mammal research and conservation. Our ultimate aim is to facilitate collab...

Lessons learned from acoustically tracking baleen whales on the Navy’s Pacific Missile Range Facility

The Journal of the Acoustical Society of America

Detection, localization, classification, and tracking of marine mammals has been performed on the... more Detection, localization, classification, and tracking of marine mammals has been performed on the U.S. Navy’s Pacific Missile Range Facility (PMRF) for over a decade. The range hydrophones are time-synchronized, have excellent spatial coverage, and monitor an area of approximately 1200 km2. Even with these ideal assets, there are hidden challenges when attempting acoustic density estimates of baleen whales on the range. This talk will cover lessons learned from tracking several species of baleen whales on the range over the last decade.

North Pacific minke whales call rapidly when calling conspecifics are nearby

Frontiers in Marine Science

North Pacific minke whale (Balaenoptera acutorostrata) boing calls are commonly detected in Hawai... more North Pacific minke whale (Balaenoptera acutorostrata) boing calls are commonly detected in Hawaiian waters. When producing boing vocalizations, minke whales seem to be in one of two calling behavioral states. Most often minke whales produce boings with inter-call intervals of several minutes, but sometimes minke whales call rapidly with inter-call intervals of less than a minute. Since minke whales are difficult to detect visually, cue-rate-based density estimation using passive acoustic monitoring has been proposed. However, the variables that influence cue rate or calling rate are poorly understood in most whales, including minke whales. We collected passive acoustic recordings from 47 bottom-mounted hydrophones at the Pacific Missile Range Facility’s instrumented range off the coast of Kauaʻi, Hawaiʻi to test the hypothesis that minke whales call more rapidly when closer in proximity to other calling conspecifics. A total of 599 days of data were recorded between August 2012 and...

Download

FY18 Annual Report on Pacific Missile Range Facility Marine Mammal Monitoring

BARSTUR Barking Sands Tactical Underwater Range BSURE Barking Sands Underwater Range Expansion BR... more BARSTUR Barking Sands Tactical Underwater Range BSURE Barking Sands Underwater Range Expansion BREVE ONR project Behavioral Response Evaluation Employing robust baselines and actual US Navy training. Award Number: N000141612859 COMPACFLT Commander Pacific Fleet CPA Closest point of approach CSM Cross Seamount is the location southwest of Hawaii where clicks with rapid intervals and frequency modulation similar to beaked whale foraging clicks were discovered. It is believed these clicks are produced by a beaked whale species other than Cuvier's or Blainville's (McDonald, 2009). DCLTDE Detection, classification, localization, tracking and density estimation laboratory located at NIWC Pacific in San Diego, California.

Download

TemporalContext-2021-FinWhale

This dataset contains representative fin whale (Balaenoptera physalus) acoustic data and annotati... more

FY16 Annual Report on PMRF Marine Mammal Monitoring

The citation of trade names and names of names of manufacturers is not to be construed as officia... more

Download

Minke whales change their swimming behavior with respect to their calling behavior, nearby conspecifics, and the environment in the central North Pacific

Frontiers in Marine Science

Behavioral responses to sonar have been observed in a number of baleen whales, including minke wh... more Behavioral responses to sonar have been observed in a number of baleen whales, including minke whales (Balaenoptera acutorostrata). Previous studies used acoustic minke whale boing detections to localize and track individual whales on the U.S. Pacific Missile Range Facility (PMRF) in Kaua ‘i, Hawai‘i before, during, and after Navy training activities. These analyses showed significant changes in central North Pacific minke whale distribution and swimming behavior during Navy sonar events. For the purposes of contextualizing changes in animal movement relative to Navy sonar, we expanded on this research to examine the natural variation in minke whale movement when Navy sonar was not present. This study included 2,245 acoustically derived minke whale tracks spanning the years 2012–2017 over all months that minke whales were detected (October–May). Minke whale movement was examined relative to calling season, day of the year, hour of day, wind speed, calling state (nominal or rapid), a...

Download

RAVEN X High Performance Data Mining Toolbox for Bioacoustic Data Analysis

arXiv (Cornell University), Oct 12, 2016

The goals of this project are [a] to develop a MATLAB toolbox, called Raven-X, intended for high ... more The goals of this project are [a] to develop a MATLAB toolbox, called Raven-X, intended for high performance data mining (detection and classification of target signals) of BIG acoustic datasets and [b] to make the toolbox freely available to the bioacoustic community. OBJECTIVES Our objective is to integrate high performance computing (HPC) technologies and bioacoustics data-mining capabilities by offering a MATLAB-based toolbox called Raven-X. Raven-X will provide a hardware-independent solution, for processing large acoustic datasets-the toolkit will be available to the community at no cost. This goal will be achieved by leveraging prior work done by Dugan, et al., [1-3] which successfully deployed MATLAB based HPC tools within Cornell University's Bioacoustics Research Program (BRP). These tools enabled commonly available multi-core computers to process data at accelerated rates to detect and classify whale sounds in large multi-channel sound archives [4, 5]. Through this collaboration, we will expand on this effort which was featured through Mathworks research and industry forums [6, 7], incorporate new cutting-edge detectors and classifiers, and disseminate Raven-X to the broader bioacoustics community. APPROACH Raven-X will provide capabilities to accelerate processing of large sound archives. The toolkit is designed to run on commercially available off the shelf (COTS) hardware (e.g. standard computers) as well as HPC technologies (e.g. cloud computers with multiple nodes). Raven-X will be designed to handle various sound formats; (e.g. flac, wav, aif, and M3R dat) and recording scenarios; such as multi-channel data and intermittent recordings. The goal is to make this tool accessible and useful for running a host of algorithms.

Download

Supplement 1: from Improve automatic detection of animal call sequences with temporal context

Full performance assessments of the various models developed in the study.

Supplement 3: from Improve automatic detection of animal call sequences with temporal context

Recall as a function of Inter-note Intervals.

Supplement 2: from Improve automatic detection of animal call sequences with temporal context

Recall as a function of Signal to Noise Ratio.

The Raven-X Software Package. A Scalable High-Performance Computing Framework In Matlab For The Analysis Of Large Bioacoustic Sound Archives

Raven-X is a software package designed for scalable high-performance computing. The software fram... more Raven-X is a software package designed for scalable high-performance computing. The software framework is written in Matlab and uses parallel-distributed computing for the analysis of large bioacoustic sound archives. This application contains various algorithms used for marine mammal sound detection. The various algorithms are available as sub-modules. These include Matched-Filter processing (DTP1D sub-module), Advanced Segmentation Recognition for tonal and pulse trains (ASR sub-module), Whistle Detection and Tracking (Silbido sub-module) and Generalized Power Law (GPL sub-module). These algorithms have a variety of pre-defined methods tuned for detecting Minke Whale, Humpback Whale, Fin Whale, Blue Whale, Elephant Rumbles, Elephant Gunshot and Mid Frequency Sonar. Users with development skills are welcome to create variations of these detectors, or interface new algorithms to the HPC software system. The full version of the software provides a user interface for operation. Operat...

Faster than real-time automated acoustic localization and call association for humpback whales on the Navy's Pacific Missile Range Facility

The Journal of the Acoustical Society of America, 2014

Optimal time difference of arrival (TDOA) methods for acoustically localizing multiple marine mam... more Optimal time difference of arrival (TDOA) methods for acoustically localizing multiple marine mammals have been applied to the data from the Navy's Pacific Missile Range Facility in order to localize and track humpback whales. Modifications to established methods were necessary in order to simultaneously track multiple animals on the range without the need for post-processing and in a fully automated way, while minimizing the number of incorrect localizations. The resulting algorithms were run with no human intervention at computational speeds faster than the data recording speed on over 40 days of acoustic recordings from the range, spanning several years and multiple seasons. Spatial localizations based on correlating sequences of units originating from within the range produce estimates having a standard deviation typically 10 m or less (due primarily to TDOA measurement errors), and a bias of 20 m or less (due to sound speed mismatch). Acoustic modeling and Monte Carlo simulations play a crucial role ...

New observations and modeling of an unusual spatiotemporal pattern of fish chorusing off the southern California coast

The Journal of the Acoustical Society of America, 2013

The purpose of this presentation is to present new results on an unusual spatiotemporal pattern o... more The purpose of this presentation is to present new results on an unusual spatiotemporal pattern of fish chorusing off the southern California coast. Characteristics of this fish chorus have been reported previously; it occurs at night in the late spring and summer months in shallow, sandy bottom regions just outside the surf zone. The background sound levels increase by up to 30 dB and cycle in level with a period of 30-35 sec all night long. In this paper, recent results from measurements made by a set of high spatial resolution sensor systems spanning a 50-km stretch of coastline out to 20 km offshore over a 2-month time period are presented. These data allow the spatial dependence and long-term temporal variability of the chorus to be examined at high spatial resolution. Refinements to a numerical model that predicts this chorusing behavior are required to account for some aspects of these new observations. [Work supported by the Office of Naval Research, Code 322-MMB].

Download

Performance metrics for marine mammal signal detection and classification

The Journal of the Acoustical Society of America

Automatic algorithms for the detection and classification of sound are essential to the analysis ... more Automatic algorithms for the detection and classification of sound are essential to the analysis of acoustic datasets with long duration. Metrics are needed to assess the performance characteristics of these algorithms. Four metrics for performance evaluation are discussed here: receiver-operating-characteristic (ROC) curves, detection-errortrade-off (DET) curves, precision-recall (PR) curves, and cost curves. These metrics were applied to the generalized power law detector for blue whale D calls [Helble, Ierley, D'Spain, Roch, and Hildebrand (2012). J. Acoust. Soc. Am. 131(4), 2682-2699] and the click-clustering neural-net algorithm for Cuvier's beaked whale echolocation click detection [Frasier, Roch, Soldevilla, Wiggins, Garrison, and Hildebrand (2017). PLoS Comp. Biol. 13(12), e1005823] using data prepared for the 2015 Detection, Classification, Localization and Density Estimation Workshop. Detection class imbalance, particularly the situation of rare occurrence, is common for long-term passive acoustic monitoring datasets and is a factor in the performance of ROC and DET curves with regard to the impact of false positive detections. PR curves overcome this shortcoming when calculated for individual detections and do not rely on the reporting of true negatives. Cost curves provide additional insight on the effective operating range for the detector based on the a priori probability of occurrence. Use of more than a single metric is helpful in understanding the performance of a detection algorithm. V

Download

Supplementary material from "Improve automatic detection of animal call sequences with temporal context

Many animals rely on long-form communication, in the form of songs, for vital functions such as m... more Many animals rely on long-form communication, in the form of songs, for vital functions such as mate attraction and territorial defence. We explored the prospect of improving automatic recognition performance by using the temporal context inherent in song. The ability to accurately detect sequences of calls has implications for conservation and biological studies. We show that the performance of a convolutional neural network (CNN), designed to detect song notes (calls) in short-duration audio segments, can be improved by combining it with a recurrent network designed to process sequences of learned representations from the CNN on a longer time scale. The combined system of independently trained CNN and long short-term memory (LSTM) network models exploits the temporal patterns between song notes. We demonstrate the technique using recordings of fin whale (<i>Balaenoptera physalus</i>) songs, which comprise patterned sequences of characteristic notes. We evaluated several variants of the CNN + LSTM network. Relative to the baseline CNN model, the CNN + LSTM models reduced performance variance, offering 9–17% increase in area under the precision–recall curve and 9–18% increase in peak F1-scores. These results show that the inclusion of temporal information may offer a valuable pathway for improving the automatic recognition and transcription of wildlife recordings.

Acoustic detection, localization, and tracking of vocalizing humpback whales on the U.S. Navy's Pacific Missile Range Facility

The Journal of the Acoustical Society of America, 2014

A subset of the 41 deep water broadband hydrophones on the U.S. Navy's Pacific Missile Range ... more A subset of the 41 deep water broadband hydrophones on the U.S. Navy's Pacific Missile Range Facility (PMRF) to the northwest of Kauai, Hawaii was used to acoustically detect, localize, and track vocalizing humpback whales as they transited through this offshore range. The focus study area covers 960 square kilometers of water (water depths greater than 300 m and more than 20 km offshore). Because multiple animals vocalize simultaneously, novel techniques were developed for performing call association in order to localize and track individual animals. Several dozen whale track lines can be estimated over varying seasons and years from the hundreds of thousands of recorded vocalizations. An acoustic model was used to estimate the transmission loss between the animal and PMRF hydrophones so that source levels could be accurately estimated. Evidence suggests a Lombard effect: the average source level of humpback vocalizations changes with changes in background noise level. Additionally, song bout duration, c...

Temporal context improves automatic recognition of call sequences in soundscape data

Journal of the Acoustical Society of America, 2020

Improving marine mammal classification using context from multiple hydrophones

Journal of the Acoustical Society of America, 2017

Detection, classification, localization, and tracking (DCLT) of marine mammals is oftentimes perf... more Detection, classification, localization, and tracking (DCLT) of marine mammals is oftentimes performed in that order. However, in the sonar-signal processing communities and elsewhere, classification is usually the final step. Thus, more appropriately, the working order should be “DLTC.” If classification is performed as the final step, the results can be greatly improved by using the context of the calls. By grouping likely calls into tracks, a collective of calls can provide much more information for classification than single calls alone. Additionally, when multiple species are calling at the same time, the location of the calls can be used to distinguish confusing signals. The time-series and spectral information of a call can also be enhanced by localizing first, and choosing the nearest hydrophone to the calling animal for signal analysis. If localization is not possible, classification can still be improved if two or more hydrophones are available with overlapping coverage, b...

The use of context in machine learning for bioacoustics

Journal of the Acoustical Society of America, 2018

Tyler Helble

Uploads

Papers by Tyler Helble

Log In