I am an electrical engineer, but not very fluent in DSPs. I have a unique application that I'm hoping someone might have a solution for.  I have a noise problem.  Literally.

In my neighborhood, there are many vehicles with modified exhaust systems.  I would like to use a microphone to record traffic noise next to a road, then use software to profile the dofferent types of vehicles, allowing me to form statistics of the frequency of certain types of vehicles (trucks, motorcycles, cars).  Ideally, I'd like to construct from these recordings a noise profile of some kind that allows a computer to identify the type of vehicle.  Ideally, I'm wondering if it's possible to identify a specific individual vehicle simply by its exhaust noise, but I'm guessing that's too specific.

Not sure if this is the right place to post such a problem, but it seemed logical.  If not, if anyone could point me in a direction that might help, I'd very much appreciate it.

Thanks

-John

Hi,

These are some quick search results you may be interested;

Classification of Vehicles Based on Audio Signals using Quadratic Discriminant Analysis and High Energy Feature Vectors, https://arxiv.org/pdf/1804.01212.pdf

Vehicle detection and classification from acoustic signal using ANN and KNN https://ieeexplore.ieee.org/document/6731694/

EXPLORING SOUND SIGNATURE FOR VEHICLE DETECTION AND CLASSIFICATION USING ANN http://www.airccse.org/journal/ijsc/papers/4213ijs...

Vehicle detection and classification from acoustic signal using ANN and KNN https://ieeexplore.ieee.org/document/6731694 

Spectral features for audio based vehicle and engine classification https://link.springer.com/article/10.1007/s10844-0...

I used to work in a research group at Ford in a group focused on sound and audio and one of the things they (before my time) was develop a a database of similar noise profiles but for different road noise conditions (i.e. different speeds on different types of terrains) among other things. If one of the big 3 hasn't done something similar I would be extremely surprised. Whether or not they published the results is another question...

In the event they did publish them, I'd check out any research journals related to automotive engineering. The SAE (Society of Automotive Engineers) is likely a good start. Regarding the noise characteristics of different vehicles you definitely want to check out the topic of NVH. https://en.wikipedia.org/wiki/Noise,_vibration,_an...

The NVH guys tune all the aspects of a car that makes noise. It's a mixture of mechanical engineering and pyschoacoustics and they're the people who do things like change the muffler's dimensions so a guy who gets a pick-up truck experiences the sound/feelings of what he previously thought a pick-up truck was supposed to sound like. These days, with electric vehicles becoming more prevalent, there's a lot of work going on in terms of sound design for engine sound synthesis.

Given that there is a safety component related to audibly being able to identify a car, I wouldn't be surprised if there were publicly available regulations which specified noise characteristics. I used to work on the "chime" issue (i.e. providing the driver audible feedback about certain things, aka making it beep when you go out of your lane) and there were legal specifications regarding those things IIRC (been awhile...).

In terms of being able to classify them, that potentially seems more like a task better suited to an ML approach, however depending on the noise characteristics (which I don't fundamentally know) a DSP approach would also work.