Across Acoustics

Nonlinear Time-warping Made Simple

January 04, 2022 ASA Publications' Office
Across Acoustics
Nonlinear Time-warping Made Simple
Show Notes Transcript

Nonlinear time-warping made simple: A step-by-step tutorial on underwater acoustic modal separation with a single hydrophone

The Journal of the Acoustical Society of America (JASA)
https://doi.org/10.1121/10.0000937

Authors: Julien Bonnel, Aaron Thode, Dana Wright, and Ross Chapman

In this episode, we interview Julien Bonnel from Woods Hole Oceanographic Institution  about time-warping.

Read more from The Journal of the Acoustical Society of America (JASA).

Learn more about Acoustical Society of America Publications.


Music Credit: Min 2019 by minwbu from Pixabay. https://pixabay.com/?utm_source=link-attribution&utm_medium=referral&utm_campaign=music&utm_content=1022 



Ambri Phillips  00:06

Welcome to Across Acoustics, the official podcast of the Acoustical Society of America’s Publications office. On this podcast, we will highlight research from our four publications, The Journal of the Acoustical Society of America, also known as JASA, JASA Express Letters, Proceedings of Meetings on Acoustics, also known as POMA, and Acoustics Today. I'm your host, Ambri Phillips, Business Administrator for the ASA.

 

Joining me today is Julien Bonnel of Woods Hole Oceanographic Institution. We will be discussing his article “Nonlinear time-warping made simple: A step-by-step tutorial on underwater acoustic modal separation with a single hydrophone,” which appeared in the March 2020 issue of JASA. Thank you for taking the time to speak with us today, Julien, how are you?

 

Julien Bonnel  01:00

Pretty good. Thanks. How are you?

 

Ambri Phillips  01:03

I'm good. Thank you. So can you provide us with a little bit of background on yourself?

 

Julien Bonnel  01:07

Sure. So I'm French. I was born in Paris in 1984. I stayed there until I was 20. After that, I went to Grenoble which is basically a French city in the middle of the Alps. I got my master degree here in 2007. That was a degree in Computer and Electrical Engineering. As part of that degree, I did a six month internship at Ryerson University in Toronto, Canada. After that, I came back to Grenoble for another three year, became a PhD student and got my PhD in 2010. I was working in signal processing and ocean acoustics. During that PhD, I did my first ASA meeting that was in 2009 in Portland in the US. I also did another internship, but this time at Scripps in San Diego in the US as well. I guess those were my first interaction with the Acoustical Society of America so that was a good time. After that I got hired as an Assistant Professor at ENSTA Bretagne in Brest, I kept working in signal processing and ocean acoustics. I've been recently hired as an Assistant Scientist at Woods Hole, so that was in 2017. I got tenured last year and I keep working in signal processing and ocean acoustics. I'm also teaching the the Array Processing class and the Ocean Acoustics class in the MIT/WHOI joint program. 

 

Ambri Phillips  02:29

That's awesome. Can you provide me with some background on underwater acoustic signal processing?

 

Julien Bonnel  02:35

Yes, so ocean acoustics is the study of how the sound propagates underwater on how we can use it to learn stuff about the ocean. Historically, that has been driven mostly by the Navy, because they were hunting for submarines. And also, that has been used a lot to study marine mammals because they do a lot of sound underwater. But today, we know that we can do much more than that. For example, we can use sound to study fish, we can use sound to study crustaceans or ice or weather or earthquakes and a lot of other things. Ocean acoustics also includes a lot of nerdy questions about the underlying physics, understanding the propagation of the sound, understanding the scattering, understanding how we can couple the physical oceanography model with the acoustic model, and how do we solve that, computationally speaking. And there is also a really important question right now, it’s the question of the noise pollution. It's now widely recognized that the noise we do at sea, we human, for example, when we take a ship, marine traffic, or oil and gas exploration or sonar system, this is doing noise. And we know that this is a pollution and it likely has an impact on the ecosystem. So we need to understand and quantify this impact. That's another important topic in ocean acoustics right now. And you also asked about signal processing. Well, when we do ocean acoustics at sea we use hydrophones which are underwater microphones, and these collect data.And the signal processing is computational methods to try to make sense of that data that we collect at sea. If you put a hydrophone at sea, you will record a cacophony with a lot of sounds. Some of them you care, some of them you don't care. So the idea behind signal processing is to extract what's important for you. It's a lot of Applied Maths, notably statistics and algebra, but it has a lot of real life application. So we use that a lot in ocean acoustics, but you know, it's used in many fields. For example, there was a lot of clever signal processing in our cell phones.

 

Ambri Phillips  04:35

Interesting. What is the difference between active acoustics and passive acoustics monitoring?

 

Julien Bonnel  04:41

So, when we do active acoustics, we use a sound source, and when we do passive acoustics we don't use a source, we just listen. So an example of active acoustics are you know, animals doing echolocation like a bat or dolphin, they will do a sound, the sound will propagate, it will bounce somewhere, and then it will come back. You know, they will use that to know what's in front of them, basically. We do exactly the same like the oil and gas company when they're looking for oil below the sea floor, they will use the same idea, or the Navy when they are hunting for submarine they use similar active acoustics sonar system. On the other hand, passive acoustics is just listening. So the Navy is doing that as well. For example, when you're in a submarine, you don't want to do any sound because you don't want people to know where you are. So if you just listen and try to infer information from what you hear. Bioacousticians that will listen for marine animals is also doing passive acoustics. Or you right now, listening to my voice, is inferring information from what you hear.

 

Ambri Phillips  05:46

Okay, that makes sense. What are hydrophones? And why are they important to underwater acoustics experiments?

 

Julien Bonnel  05:55

So a hydrophone is basically just an underwater microphone. And it's important because we need experimental data. And you know, that just requires dedicated equipment that survives underwater.

 

Ambri Phillips  06:08

Okay. What is warping? And how does it work?

 

Julien Bonnel  06:13

So warping is a signal processing trick to maximize both the amount and the quality of information that we can extract from data collected using a single hydrophone. That is important because usual ocean acoustic experiments, they use arrays. So we try to do stuff with a single hydrophone and that obviously radically simplify the burden, the time needed, and the cost associated with ocean acoustics experiment.

 

Ambri Phillips  06:42

Okay, and how would one go about implementing warping?

 

Julien Bonnel  06:46

Okay, so that's obviously a bit of a tricky question, but I'll try to make that not overly complicated and boring. Some of you may know about Star Trek, or any sci fi movie, where they try to travel faster than light. Basically, what they do is that they warp space, they change the shape of space, so that they can take a shoter pass. And, you know, from a mathematical point of view, if you can warp space, you can also warp time. Obviously, you cannot do that in, you know, in real life. But if you've got data that is recorded on your computer, you can do it on a computer, for example, you know, with an old tape player machine, you can play faster, right? And basically we do the same with digital data, we do the same with signal processing methods. And that's what we do, we change the way the signal is sampled, and if we do that in a clever way that can be helpful to extract information.

 

Ambri Phillips  07:49

Okay, great. That makes sense. Um, what are some practical considerations when warping?

 

Julien Bonnel  07:56

I told you that warping is stretching time ... and the question is, you know, how do we stretch? Like, do I stretch a lot? Do I stretch a bit? In which direction? And to do that, we need to know the underlying physics, we need to have like models of what's going on. And for us, that is the physics of sound propagation at sea. So obviously, we need hypothesis, and we need that hypothesis to be not too wrong so that warping works. And practically speaking, that warping trick works mostly in coastal shallow water. So that means you know, when water that is less than a few hundred meters, for relatively low frequency sound, so that means when the frequency is less than a few hundreds hertz, you know, mid to long range when the source receiver is more than a few kilometers. And this is important because that put us in a regime where the propagation is specific and we have good model. So that we can stretch our signal in a way that is meaningful with respect to the underlying propagation. It also requires the source to be relatively brief, you know, something like a clap *claps* like that, or maybe an explosion or a marine mammal vocalization or a man-made tomographic sound source. It does not work if the source is too long, something like continuous ship noise, that will not work.

 

Ambri Phillips  09:14

And what are some applications of warping?

 

Julien Bonnel  09:18

So warping allows to localize sound sources, and also to estimate the environmental property between the source and the receiver. And maybe I can give you a few examples. We've demonstrated that we can use warping to localize baleen whales, the large one like the blue whale, for example, from the sound they make. And we notably applied that to study the North Pacific right whales and the bowhead whales in the Arctic. Warping can also be used to estimate the property of the environment between the source and the receiver. You can say I'm listening to a source I don't really care about that source, but I want to estimate the environment between that source and me. And that is very important because you need this environmental information to properly predict the propagation in an area of interest; and predicteing the propagation is important, for example, for the Navy to be able to assess sonar performances, but also for stakeholders if they want to monitor or forecast noise pollution they need to know this information, and we can use warping to get that. 

 

 

Ambri Phillips  10:20

Okay. Are there any other examples of warping experiments?

 

 

Julien Bonnel  10:25

Yep, yeah. So, you know, I think that the beauty of warping is that it makes experiments at sea so much simpler, because you can deploy a single hydrophone instead of a large array. For example, we recently perform a test and show that we can do our traditional geoacoustic experiments, you know, from a fishing vessel instead of a large vessel result. So that was cool. Another thing is that, we can also use warping to rethink the way we do acoustic experiments. For example, next year, as part of the Seabed Characterization Experiment, I will deploy a set of 20 hydrophones over an area of interest. So that will be some kind of, you know, an acoustic carpet, as I like to call it. And this is important, because using warping, we'll be able to use that carpet to fully characterize the spatial variability of the seafloor in this area. And if we were to do the same without warping, instead of 20 single hydrophones, we would need 20 arrays, that would be just totally crazy. You know, even if you have enough money to get these arrays in the water, just the deployment and recovery time would be prohibitive. And one nice thing about warping is that it also allows to revisit old dataset and to extract more information than what was initially planned. Examples include, you know, there are many, many bioacoustic dataset in the lab. Usually they have been collected with a single hydrophone. They've been studied to detect the presence of marine mammal and using warping we can revisit these data sets and not only detect the animal but also localize them. And that's really important for the biologists.

 

Ambri Phillips  12:05

Interesting. Is there any additional information that we should know about time warping?

 

Julien Bonnel  12:10

Sure. Well, it's super fun. On the why, why is it fun? Well, it's fun because, you know, it requires a lot of skills and we touched most of them but you know, it's some kind of signal processing. Also, you need to understand the basic physics and you need to understand the application on you know, work with many people including, you know, biologist, acousticians, and signal processor and that  makes it fun, at least to me.

 

Ambri Phillips  12:37

Okay, I understand. Well, thank you for taking the time to speak with us today. I'm sure our listeners will be happy to have been provided with more insight on time-warping. 

 

Thank you for tuning in to Across Acoustics. If you'd like to hear more interviews from authors about their research, hit subscribe and find us on your preferred podcast platform.