noise pollution

[gusl]

Stichting BAM: Bescherming Akoestisch Milieu (Protection of the Acoustic Environment) for people who are highly sensitive to noise. Their goal is to fight "imposed noise".

They do activism and have information on quiet holiday resorts, and I would suggest make guides to quiet places to go out in the city. Most of the members are musicians, writers or journalists, and it turns out that Pamela Hemelrijk (libertarian journalist) and Paul Vitanyi (who wrote the Kolmogorov Complexity book) are among them.

A high-tech libertarian solution for this might be for each of us to adjust our own sensitivity... i.e. "standardize", but I think our technology isn't good enough to make wearing the appropriate equipment comfortable.

Another solution would be for sound producers to only broadcast to the intended recipients. Directional sound is a big step in the right direction. You-know-who-you-are: you've got your work cut out for you.

---

What's the difference between The Mute and noise-cancelling earphones? The Mute samples the undesired noise, but they can hardly assume that children screaming makes a constant sound.

Comments

[zarex.livejournal.com]

Yup.

The Mute is basically the same as noise-cancelling headset, although they have a different hype-angle, since they're presumably looking for investors. And you're right - most sounds you want to block aren't constant nor predictable, which is why most noise cancellation (except in headphones) doesn't work very well. It's a prediction problem, not a reproduction problem.

[gustavolacerda.livejournal.com]

http://www.livejournal.com/users/gustavolacerda/280215.html?thread=905111#t905111

how smart can you get with the filtering? I can imagine that we could do better than just frequency filtering... I think they could do better smart filtering with a sample of the undesirable noises.

[zarex.livejournal.com]

It has to be exactly cancelling, otherwise you'll just get more noise than you started with. Predictive models can be used, but they aren't going to help much, since error comes at a high penalty, and you don't have any lag to predict against anyhow.

I think it's all marketing hype.

[gustavolacerda.livejournal.com]

Well, the lag is the time is takes for the sound to travel from the mic outside the headphone to the ear. Doesn't noise-cancellation work inside this very small margin, or does it work on a continuity assumption (i.e. "the next 10 milliseconds will probably be similar to the last 10 milliseconds")?

Theoretically it's straightforward, as long as the undesirable noises are learnable... there are definite intonation patterns in children's screaming.

[zarex.livejournal.com]

The lag is small enough to be insignificant (in fact, it might be long enough to require the cancelling signal to be delayed). Sound only moves about 1ft/ms. For a frequency of 1kHz, the wavelength is 1 ft, so there's plenty of time/space to cancel low frequencies.

Undesirable sounds aren't learnable (not well, anyhow). It's a really, really hard problem. Harder even to cancel them, since you'd need them to be exact.

[ernunnos.livejournal.com]

I imagine they're just taking a spectrogram of the sound you record, and then using that as a template for cancellation in the future. General noise-cancellation headphones just cancel everything below around 1500 Hz. The systems aren't precise/fast enough to cancel anything above that, and that tends to hit a lot of very annoying noise anyway, so it's good enough. I highly doubt "The Mute" is any better at the actual cancellation, so it seems mostly just a marketing ploy.

Now what you could do is a combination of passive and active noise cancellation. Use over-ear earmuffs with microphones on them. This gives you a better environment for active noise cancellation, and kills the high frequencies too. Then selectively allow frequencies either by not cancelling them (< 1500 Hz) or by simply playing them back through the headphones after running them through an attenuating filter (> 1500 Hz).

My shooting headphones actually work similarly, blocking everything passively, and passing through selectively with amplification & compression. (Loud sounds aren't included in the feed, quiet sounds are amplified.)

[gustavolacerda.livejournal.com]

I imagine they're just taking a spectrogram of the sound you record, and then using that as a template for cancellation in the future.

Right.. I thought that it might use this information to recognize undesirable sounds, and filter them while trying not to filter out good sounds. In that sense it would be better than regular noise-cancelling headphones.


shooting headphones
Is that something you made?

[zarex.livejournal.com]

Lots of noise cancelling headphones do just what ernunnos mentions - passive filtering where it's effective (high frequencies) and active cancellation where it's effective (low freqs). It's a good design choice.

Personally, I like my Eytmotic plugs, which are all passive, but they're comfortable, and very effective in blocking sound. And they're the best sounding headphones I've ever heard.

The template idea makes some sense, but acoustic source separation is a devilishly difficult, unsolved problem in signal processing, so I doubt they'll make much improvement. Frequency filtering by "good source" template is a rudimenary approach, and it's probably better than nothing ... but not by much.

[gustavolacerda.livejournal.com]

acoustic source separation is a devilishly difficult, unsolved problem in signal processing, so I doubt they'll make much improvement

I know this, since there is no system to automatically transcribe polyphonic music... but I've always wondered why!!
It seems so straightforward: just separate the peaks (along with their multiples) in the Fourier transform... what's the problem with this approach??

[zarex.livejournal.com]

The number of dissertations written on this stuff is staggering; it's a lot harder than it appears. Even transcribing monophonic melodies is tough (with real instruments), but it's doable.