Proposal talk:Objects generating audible cues

Broadcast RFC

• https://lists.openstreetmap.org/pipermail/tagging/2020-September/055509.html

Intrigued

I'm intrigued by this proposal. At a glance, I view it as an analogue to colour=* but for aural tagging: as mappers, we don't prioritize tagging everything with its color, but sometimes the color is unusual enough to warrant an explicit tag. So too, sometimes the sound is unusual. There's an entire genre of public artwork that would qualify for this tag.

I've long wanted to make a "renderer" on iOS that would take advantage of SceneKit's positional audio functionality. In general, I would've expected to come up with heuristics based on existing feature tags, such as a fluttering sound for a man_made=flagpole when simulating a windy day, or a wailing emergency=siren on a stormy day. This proposal would make it possible to provide more detail or override a heuristic to reflect an exception, such as a siren model that supports modes beyond the usual "wail" and "attack", or perhaps a musical road.

That said, I'm not familiar enough with the vocabulary around aural mapping to know whether the specific tags in this proposal are appropriate, universal, and specific enough. Are these long lists of tags merely suggestions of potentially common tags, with mappers coining additional tags as needed, or are they intended to be more or less comprehensive? In addition to the proposed audible:spl=* tag, should be there be a generalization of the existing siren:range=* tag for other noise generators?

 – Minh Nguyễn ^💬 20:34, 9 October 2020 (UTC)

Thank you for your insights. Could you perhaps list audible artworks? Good catch about the musical roads - I forgot about them and it would again fit this concept, although the visual impaired aren't usually driving cars, so it probably wouldn't be that useful for them (however, this would still be interesting for the entertainment value to drivers).

I think the micromapping of the kind of sonification that an emergency siren features is more similar to specifying how traffic signals talk or whine and may perhaps go within siren:sound=* or something similar. Although, maybe I'm misunderstanding - if such a siren can be reliably heard all day long and used for navigation purposes, we should reconsider.

Let me sleep on your question about range to come up with a better answer. It is pretty straight forward to measure or estimate dBA using your smartphone, but range can get tricky. I acknowledge that according to my aural mapping experience, certain objects can be heard from further away than others if they are taller and/or placed higher above ground. I try to indicate this in height=* and min_height=*, but nearby vegetation and objects can greatly impact range and this is not straightforward to model. High-rise objects also suffer from refraction in certain conditions (inversion) that could greatly alter its range and generally varies with temperature and humidity, so it is not an easily verifiable property. Also most non-cylindrical objects seem to be highly anisotrophic - their volume is in the same ballpark in most directions, but certain directions towards corners are clearly less loud than at other places. As I mostly recommend using audible:spl=* as a ballpark figure to indicate whether something is faint, normal, moderate or excessively loud. If something is very faint, it is easy to miss in traffic for example.

The tag suggestions here have been made based on my observation and mapping experience in my area. It would be great if we could enumerate as many common noise sources as possible to be able to come up with a better naming convention and possible subtags - I welcome all of your suggestions. This is how I decided to decomposed the original ideas of humming/buzzing/whining into tone + fundamentals=xxx. I don't think that this list can be any shorter, but making it completely freeform or listing too many similar choices would make interpretation more and more difficult. I was also thinking about inserting samples as "screenshots" as well. -Bkil (talk) 23:45, 9 October 2020 (UTC)

If the goal is to develop a detailed, reasonably comprehensive set of tags, it may be prudent to base the tags on an existing ontology along the lines of AudioSet. Unfortunately, the AudioSet ontology is licensed under the Creative Commons Attribution-ShareAlike license, but perhaps there are unencumbered alternatives out there. In any case, these ontologies may serve as inspiration for making the set of suggested tags more diverse and versatile. – Minh Nguyễn ^💬 15:47, 10 October 2020 (UTC)

So, did you have any concrete recommendation in mind based on that ontology? I've looked through all of their categories but couldn't find anything missing that I could plausible find out on the streets. Although, it would be great if you could find an alternative to what I at present refer to as "murmur", because I think it's not the proper word. These are low frequency dominant "calm", non-water like sounds heard at certain manhole covers or conducted by water pipes towards taps or fire hydrants. Again, have you encountered anything around your area that gives out steady sounds useful for navigation? I've uploaded almost a hundred until now, but I would be interested what would be missing. -Bkil (talk) 18:44, 13 October 2020 (UTC)

Guard dogs

It has been raised on a private discussion if we could mention whether a private house or an industrial site has a guard dog that is easily identifiable by its barking. It is my viewpoint that from a mapping ethics standpoint, we should not map this because it may compromise a home's security. -Bkil (talk) 14:17, 14 October 2020 (UTC)

Have anyone tried using this tagging

Have anyone tried using this tagging? I am skeptical about usefulness and "This key does not appear in the OSM database." suggests that noone tested this proposal on even a small scale Mateusz Konieczny (talk) 08:55, 5 October 2020 (UTC)

Good question. Yes, I have been experimenting with in the last few weeks. I've collected dozens of various audible things as locally marked map notes. If you think it would help, I'd gladly upload it to OSM. -Bkil (talk) 09:04, 5 October 2020 (UTC)

Yes, it would help. Personally I think that while making proposal for tags aleady used widely is rarely useful I am not fan of tagging proposals done without any mapping done with given scheme. If it is a tested proposal with some sample data in OSM I would be far more likely to support it (with possible exception for someone else noticing some issues with it, I see no problems with it) Mateusz Konieczny (talk) 09:24, 5 October 2020 (UTC)

Okay, I'll upload these to OSM today then. I agree that you have a better "feel" to it after you have taken the time to look for it. This is where the desire to map audible:spl=* came from - I found lots of transformers that are borderline while others can be heard from across the street. I've also just recently found the loudest fire hydrant ever - it can be heard from about 50 meters in calm conditions and it is literally frightening if you aren't expecting it. -Bkil (talk) 09:32, 5 October 2020 (UTC)

Fire hydrant making noice (while not in use), really? Do you suggest to tag only unusual sounds? In the proposal I see only the opposite (objects usually noisy but exceptionally less noisy than the average item of this kind).--Nospam2005 (talk) 19:55, 25 October 2020 (UTC)

Yes, if you mapped enough, you will have encountered both fire hydrants and drinking water taps that make a distinctive sound from across the street! This schema allows you to tag any kind of sound that could aid in navigation. I'm not sure what you mean by "exceptional", because for example part of huge power transformers I have encountered are completely silent, while another part of them can be heard from across the street. Some for telco street cabinets, though high street ones seem to produce a lot of noise in 90% of the cases. If by "exceptional" you mean "unintentional", then yes, we can say that. As outlined in the list at the end, the things that allow accessibility by purpose already have a pretty good scheme of their own. -Bkil (talk) 20:35, 25 October 2020 (UTC)

I've progressed with mapping a few areas and proceed to learn from my experience: https://overpass-turbo.eu/s/YTG Let's have at least one more week for comments - we've started discussing this with a few people and I anticipate some great insights. -Bkil (talk) 23:53, 9 October 2020 (UTC)

How do you measure the Sound Pressure Level

How do you measure the Sound Pressure Level? Do you carry a calibrated directional measurement microphone? --—M!dgard [ talk ] 10:39, 26 October 2020 (UTC)

If you search for "sound pressure level" on eBay you can find many offers starting from ~$15. You could also find smartphone apps which after calibration should also produce sufficient estimates. You will also need a distance measurement aid that can be as simple as a 1 meter long stick, or an ultrasonic or laser-based measurement instrument. After a little practice, you could easily classify loudness without instruments (I've added a chart with some example SPL for starters). Anyway, as the perceived volume of a transformer could vary with humidity, temperature and load, I wouldn't want to be overdoing it. Basically, we want to share in an objective way whether something is faint (only perceptible without traffic and/or of you are looking for it), moderate (clearly noticeable) or loud (can be heard from across the street). We may tag the measurement instrument and expected accuracy on the nodes as well if we wanted to (like audible:spl:instrument=* for a model number and/or audible:spl:accuracy=* for an expected accuracy in dBA), but I feel that to be a little overdoing it compared to what benefit it would provide to users. I originally wanted a simple 3 or 4 level list for loudness similar to road smoothness, but subjective scales are always harder to accept, accommodate and train. -Bkil (talk) 19:35, 27 October 2020 (UTC)

Cool, you should definitely document that on the wiki page! But if I get this correctly, on a busy street you would barely hear 85dB whereas in a quiet village it would be a significant auditive landmark. How will this be intelligble for data users? —M!dgard [ talk ] 13:18, 28 October 2020 (UTC)

Yes, you are right. At the moment I'm not focusing mapping of busy streets just for this reason (people usually don't like to hike there anyway). Data users should cross reference outside databases that contain point measurements for background noise levels and traffic noise charts by time of day. A 85dB point source could be heard from the same distance in a city at 4am as in a village, so its tag is still valid, but it needs to be interpreted differently based on time. I haven't done measurements on busy streets, but I think they should be done at night by default (or first measure an ambient SPL and than subtract that from the noise emitter's level, though I think that's not a very good approximation). Maybe we could have a day measurement along with the default night measurement, like audible:spl:day=85, although we could mind as well simply use the already established conditional restrictions system to tag this. -Bkil (talk) 16:56, 28 October 2020 (UTC)

Radius of audibility vs. SPL

How about tagging the distance (radius) of audibility, instead? It's a related thing, it would be useful for sonification, and possibly easier for people to tag without needing specialist equipment. Something like "audible:distance=30 m". --Danstowell (talk) 15:40, 28 October 2020 (UTC)

SPL is a standard unit of loudness measured at a distance of 1 meter. I don't think that anything else could work, in the sense that it must be verifiable (with measurements repeatable by multiple validating mappers) and convey information relating to the object at hand.

• The range that you can hear something from highly depends on your genetics and ear health,
• The background noise level (traffic, leaves, wind and whatnot),
• Temperature, humidity and wind/air pressure conditions (see inversion),
• And the exact heading of the spot that you are testing - as mentioned earlier, most noise emitters are highly anisotropic meaning that their detection range is not a circle.
• Even worse, in our 3D real world with various random occlusions, vegetation and height, it could very well happen that you can hear something from up to 10 meters away, but then after following a path away you can again hear it 100 meters further down the road as the bushes permit.
• I've just today found a power station that you can hear from at least 200 meters away but only after you know what to look for and only on a given section of road in the proper direction after a given distance - otherwise it's not even audible next to the fence of the property, maybe 50 meters away from the emitter.
• All of these reasons imply orders of magnitudes of practical vs. mapped detection range differences, which makes this field hardly useful. If what you mean by audible range is again a simple scale, like audible:range=1m=near/10m=intermediate/100m=far(across street), then that could be simply cross referenced to a dBA table on the UI. -Bkil (talk) 16:47, 28 October 2020 (UTC)