Dr. Doolittle uses AI: Ethical challenges of trying to speak whale

ABSTRACTHere we report on a rare and opportunistic acoustic “conversation” with an adult female humpback whale, known as Twain, in Southeast Alaska. Post hoc acoustic and statistical analyses of a 20-minute acoustic exchange between the broadcast of a recorded contact call, known as a ‘whup/throp’, with call responses by Twain revealed an intentional human-whale acoustic (and behavioral) interaction. Our results show that Twain participated both physically and acoustically in three phases of interaction (Phase 1: Engagement, Phase 2: Agitation, Phase 3: Disengagement), independently determined by blind observers reporting on surface behavior and respiratory activity of the interacting whale. A close examination of both changes to the latency between Twain’s calls and the temporal matching to the latency of the exemplar across phases indicated that Twain was actively engaged in the exchange during Phase 1 (Engagement), less so during Phase 2 (Agitation), and disengaged during Phase 3...

ArXiv, 2021

The past decade has witnessed a groundbreaking rise of machine learning for human language analysis, with current methods capable of automatically accurately recovering various aspects of syntax and semantics — including sentence structure and grounded word meaning — from large data collections. Recent research showed the promise of such tools for analyzing acoustic communication in nonhuman species. We posit that machine learning will be the cornerstone of future collection, processing, and analysis of multimodal streams of data in animal communication studies, including bioacoustic, behavioral, biological, and environmental data. Cetaceans are unique non-human model species as they possess sophisticated acoustic communications, but utilize a very different encoding system that evolved in an aquatic rather than terrestrial medium. Sperm whales, in particular, with their highly-developed neuroanatomical features, cognitive abilities, social structures, and discrete click-based encod...

2003

˜The œJournal of the Acoustical Society of America/˜The œjournal of the Acoustical Society of America, 2024

The study of humpback whale song using passive acoustic monitoring devices requires bioacousticians to manually review hours of audio recordings to annotate the signals. To vastly reduce the time of manual annotation through automation, a machine learning model was developed. Convolutional neural networks have made major advances in the previous decade, leading to a wide range of applications, including the detection of frequency modulated vocalizations by cetaceans. A large dataset of over 60 000 audio segments of 4 s length is collected from the North Atlantic and used to fine-tune an existing model for humpback whale song detection in the North Pacific (see Allen, Harvey, Harrell, Jansen, Merkens, Wall, Cattiau, and Oleson (2021). Front. Mar. Sci. 8, 607321). Furthermore, different data augmentation techniques (time-shift, noise augmentation, and masking) are used to artificially increase the variability within the training set. Retraining and augmentation yield F-score values of 0.88 on context window basis and 0.89 on hourly basis with false positive rates of 0.05 on context window basis and 0.01 on hourly basis. If necessary, usage and retraining of the existing model is made convenient by a framework (AcoDet, acoustic detector) built during this project. Combining the tools provided by this framework could save researchers hours of manual annotation time and, thus, accelerate their research.

Our long-range objective is to understand the oceanographic processes that influence the distribution of whales in the ocean. In support of this objective, we seek to develop new techniques and technologies that enable us to relate the occurrence and movement of animals to physical, biological, and possibly anthropogenic forcing mechanisms over long time scales. This work will ultimately improve our ability to predict whale distribution and bolster efforts to mitigate human impacts on marine mammals. Our research has three specific objectives: (1) Develop a platform-independent modular acoustic package capable of automated detection and classification of whale vocalizations and suitable for use on a variety of autonomous platforms; (2) Characterize the efficacy of several automated detector algorithms using a rich set of collocated visual and acoustic measurements collected in 2006 and 2007; and (3) Perform quantitative field trials to evaluate baleen whale detection performance in ...

The Journal of the Acoustical Society of America

Acoustic methods are becoming increasingly common in the study of marine mammal populations and behavior. Automating the detection and classification of whale vocalizations has been a central aim of these methods. The focus has primarily been on intra-species detection and classification, however, humpback whale (Megaptera novaeangliae) social call detection and classification has largely remained a manual task in the bioacoustics community. To automate this process, we processed spectrograms of calls using PCA-based and connected-component-based methods, and derived features from relative power in the frequency bins of these spectrograms. We then used these features to train and test a supervised Hidden Markov Model (HMM) algorithm to investigate classification feasibility.

The Journal of the Acoustical Society of America, 2014

Vocal communication is a primary communication method of killer and pilot whales, and is used for transmitting a broad range of messages and information for short and long distance. The large variation in call types of these species makes it challenging to categorize them. In this study, sounds recorded by audio sensors carried by ten killer whales and eight pilot whales close to the coasts of Norway, Iceland, and the Bahamas were analyzed using computer methods and citizen scientists as part of the Whale FM project. Results show that the computer analysis automatically separated the killer whales into Icelandic and Norwegian whales, and the pilot whales were separated into Norwegian long-finned and Bahamas short-finned pilot whales, showing that at least some whales from these two locations have different acoustic repertoires that can be sensed by the computer analysis. The citizen science analysis was also able to separate the whales to locations by their sounds, but the separation was somewhat less accurate compared to the computer method.

The Journal of the Acoustical Society of America, 2013

Following a production-based approach, this paper deals with the acoustic behavior of humpback whales. This approach investigates various physical factors, which are either internal (e.g., physiological mechanisms) or external (e.g., environmental constraints) to the respiratory tractus of the whale, for their implications in sound production. This paper aims to describe a functional scenario of this tractus for the generation of vocal sounds. To do so, a division of this tractus into three different configurations is proposed, based on the air recirculation process which determines air sources and laryngeal valves. Then, assuming a vocal function (in sound generation or modification) for several specific anatomical components, an acoustic characterization of each of these configurations is proposed to link different spectral features, namely, fundamental frequencies and formant structures, to specific vocal production mechanisms. A discussion around the question of whether the whale is able to fully exploit the acoustic potential of its respiratory tractus is eventually provided. V

2022

Male humpback whales (Megaptera novaeangliae) engage in complex singing displays during the winter breeding season. This hierarchically structured song is composed of individual units arranged into phrases and themes that are produced for usually 12-18 minutes at a time and often repeated for multiple hours. Singing is presumed to be important for breeding, although its exact function remains a topic of debate. The effect of rising levels of anthropogenic noise in the ocean on humpback whale singing is still poorly understood. Here we report an incident where a singer instrumented with an acoustic tag was opportunistically exposed within a few hundred meters to noise from a transiting tugboat towing a barge between islands in Hawaii. The singing whale was recorded on the tag for several song cycles before, during and after exposure to the noise event. The recordings reveal that the whale persisted in singing normally at the beginning of the exposure, notably changed the unit and phrase structure while the vessels approached, and abruptly interrupted its singing when the vessels were closest. The whale returned to normal singing once the vessel noise abated. These results shed additional light on how humpback whales respond to anthropogenic noise.

2007

This paper describes the process of creating a large digital archive of killer whale or orca vocalizations. The goal of the project is to digitize approximately 20000 hours of existing analog recordings of these vocalizations in order to facilitate access to researchers internationally. We are also developing tools to assist content-based access and retrieval over this large digital audio archive. After describing the logistics of the digitization process we describe algorithms for denoising the vocalizations and for segmenting the recordings into regions of interest. It is our hope that the creation of this archive and the associated tools will lead to better understanding of the acoustic communications of Orca communities worldwide.