Université Grenoble Alpes (UGA)
11 rue des Mathématiques
38402 Saint Martin d’Hères (GRENOBLE)
Popular version of paper 3pAB4 “Automatic fish sounds classification”
Presented Wednesday afternoon, May 25, 2016, 2:15 in Salon I
171st ASA Meeting, Salt Lake City
In the current context of global warming and environmental concern, we need tools to evaluate and monitor the evolution of our environment. The evolution of animal populations is of a special concern in order to prevent changes of behaviour under environmental stress and to preserve biodiversity. Monitoring animal populations however, can be a complex and costly task. Experts can either (1) monitor animal populations directly on the field, or (2) use sensors to gather data on the field (audio or video recordings, trackers, etc.) and then process those data to retrieve knowledge about the animal population. In both cases the issue is the same: experts are needed and can only process limited quantity of data.
An alternative idea would be to keep using the field sensors but to build software tools in order to automatically process the data, thereby allowing monitoring animal populations on larger geographic areas and for extensive time periods.
The work we present is about automatically monitoring fish populations using audio recordings. Sounds have a better propagation underwater: by recording sounds under the sea we can gather loads of information about the environment and animal species it shelters. Here is an example of such recordings:
Legend: Raw recording of fish sounds, August 2014, Corsica, France.
Regarding fish populations, we distinguish four types of sounds that we call (1) Impulsions, (2) Roars, (3) Drums and (4) Quacks. We can hear them in the previous recording, but here are some extracts with isolated examples:
Legend: Filtered recording of fish sounds to hear Roar between 5s and 13s and Drums between 22s to 29s and 42s to 49s.
Legend: Filtered recording of fish sounds to hear Quacks and Impulsions. Both sounds are quite short (<0.5s) and are heard all along the recording.
However, to make a computer automatically classify a fish sound into one of those four groups is a very complex task. A simple or intuitive task for humans is often extremely complex for a computer, and vice versa. This is because humans and computers process information in different ways. For instance, a computer is very successful at solving complex calculations and at performing repetitive tasks, but it is very difficult to make a computer recognize a car in a picture. Humans however, tend to struggle with complex calculations but can very easily recognise objects in images. How do you explain a computer ‘this is a car’? It has four wheels. But then, how do you know this is a wheel? Well, it has a circular shape. Oh, so this ball is a wheel, isn’t it?
This easy task for a human is very complex for a machine. Scientists found a solution to make a computer understand what we call ‘high-level concepts’ (recognising objects in pictures, understanding speech, etc.). They designed algorithms called Machine Learning. The idea is to give a computer a lot of examples of each concept we want to teach it. For instance, to make a computer recognise a car in a picture, we feed it with many pictures of cars so that it can learn what a car is, and with many pictures without cars so that it can learn what a car is not. Many companies such as Facebook, Google, or Apple use those algorithms for face recognition, speech understanding, individualised advertisement, etc. It works very well.
In our work, we use the same technics to teach a computer to recognize and automatically classify fish sounds. Once those sounds have been classified, we can study their evolutions and see if fish populations behave differently from place to place, or if their behaviours evolve with time. It is also possible to study their density and see if their numbers vary through time.
This work is of a particular interest since to our knowledge, we present the first tool to automatically classify fish sounds. One of the main challenges is to make a sound understandable by a computer,that is to find and extract relevant information in the acoustic signal. By doing that, it gets easier for the computer to understand similarities and differences between all signals and in the end of the day, to be able to predict to which group a sound belongs.
Many marine and aquatic human activities generate underwater noise and can have potentially adverse effects on the underwater acoustical environment. For instance, loud sounds can affect the migratory or other behavioral patterns of marine mammals  and fish . Additionally, if the noise is loud enough, it could potentially have physically damaging effects on these animals as well.
Examples of human activities that that can generate such noise are offshore wind farm installation and operation; bridge and dock construction near rivers, lakes, or ports; offshore seismic surveying for oil and gas exploration, as well as oil and gas production; and noise in busy commercial shipping lanes near environmentally sensitive areas, among others. All of these activities can generate noise over a broad range of frequencies, but the loudest components of the noise are typically at low frequencies, between 10 Hz and about 1000 Hz, and these frequencies overlap with the hearing ranges of many aquatic life forms. We seek to reduce the level of sound radiated by these noise sources to minimize their impact on the underwater environment where needed.
A traditional noise control approach is to place some type of barrier around the noise source. To be effective at low frequencies, the barrier would have to be significantly larger than the noise source itself and more dense than the water, making it impractical in most cases. In underwater noise abatement, curtains of small freely rising bubbles are often used in an attempt to reduce the noise; however, these bubbles are often ineffective at the low frequencies at which the loudest components of the noise occur. We developed a new type of underwater air-filled acoustic resonator that is very effective at attenuating underwater noise at low frequencies. The resonators consist of underwater inverted air-filled cavities with combinations of rigid and elastic wall members. They are intended to be fastened to a framework to form a stationary array surrounding an underwater noise source, such as the ones previously mentioned, or to protect a receiving area from outside noise.
The key idea behind our approach is that our air-filled resonator in water behaves like a mass on a spring, and hence it vibrates in response to an excitation. A good example of this occurring in the real world is when you blow over the top of an empty bottle and it makes a tone. The specific tone it makes is related to three things: the volume of the bottle, the length of its neck, and the size of the opening. In this case, a passing acoustic wave excites the resonator into a volumetric oscillation. The air inside the resonator acts as a spring and the water the air displaces when it is resonating acts as a mass. Like a mass on a spring, a resonator in water has a resonance frequency of oscillation, which is inversely proportional to its size and proportional to its depth in the water. At its resonance frequency, energy is removed from the passing sound wave and converted into heat through compression of the air inside the resonator, causing attenuation of the acoustic wave. A portion of the acoustic energy incident upon an array of resonators is also reflected back toward the sound source, which reduces the level of the acoustic wave that continues past the resonator array. The resonators are designed to reduce noise at a predetermined range of frequencies that is coincident with the loudest noise generated by any specific noise source.
Underwater photograph of a panel array of air-filled resonators attached to a framework. The individual resonators are about 8 cm across, 15 cm tall, and open on the bottom. The entire framework is about 250 cm wide and about 800 cm tall.
We investigated the acoustic properties of the resonators in a set of laboratory and field experiments. Lab measurements were made to determine the properties of individual resonators, such as their resonance frequencies and their effectiveness in damping out sound. These lab measurements were used to iterate the design of the resonators so they would have optimal acoustic performance at the desired noise frequencies. Initially, we targeted a resonance frequency of 100 Hz—the loudest components of the noise from activities like marine pile driving for offshore wind farm construction are between 100 Hz and 300 Hz. We then constructed a large number of resonators so we could make arrays like the panel shown in the photograph. Three or four such panels could be used to surround a noise source like an offshore wind turbine foundation or to protect an ecologically sensitive area.
The noise reduction efficacy of various resonator arrays were tested in a number of locations, including a large water tank at the University of Texas at Austin and an open water test facility also operated by the University of Texas in Lake Travis, a fresh water lake near Austin, TX. Results from the Lake Travis tests are shown in the graph of sound reduction versus frequency. We used two types of resonator—fully enclosed ones called encapsulated bubbles and open-ended ones (like the ones shown in the photograph). The number or total volume of resonators used in the array was also varied. Here, we express the resonator air volume as a percentage relative to the total volume of the array framework. Notice, our percentages are very small so we don’t need to use much air. For a fixed percentage of volume, the open-ended resonators provide up to 20 dB more noise reduction than the fully encapsulated resonators. The reader should note that noise reduction of 10 dB means the noise levels were reduced by a factor of three. A 30 dB reduction is equivalent to the noise be quieted by a factor of about 32. Because of the improved noise reduction performance of the open-ended resonators, we are currently testing this type of resonator at offshore wind farm installations in the North Sea, where government regulations require some type of noise abatement to be used to protect the underwater acoustic environment.
Sound level reduction results from an open water experiment in a fresh water lake. Various types of air-filled resonators were tested including fully encapsulated resonator and open-ended resonators like the ones shown in the photograph. Because a much total volume (expressed as a percentage here) is needed, the open-ended resonators are much more efficient at reducing underwater noise.
 W. John Richardson, Charles R. Greene, Jr., Charles I. Malme, and Denis H. Thomson, Marine Mammals and Noise (Academic Press, San Diego, 1998).
 Arthur Popper and Anthony Hawkins (eds.), The Effects of Noise on Aquatic Life, Advances in Experimental Medicine and Biology, vol. 730, (Springer, 2012).
Kevin M. Lee – email@example.com
Andrew R. McNeese – firstname.lastname@example.org
Applied Research Laboratories
The University of Texas at Austin
Preston S. Wilson – email@example.com
Mechanical Engineering Department and Applied Research Laboratories
The University of Texas at Austin
Mark S. Wochner – firstname.lastname@example.org
Popular version of paper 3aPA8
Presented Wednesday Morning, October 29, 2014
168th Meeting of the Acoustical Society of America, Indianapolis, Indiana