Hi,

this morning I had a cup of coffee outside and tried to have [sigmund~] catch on to the birds singing. My own whistling was copied perfectly, but I only got a few hints of the bird song.

I tried many different settings of -hop and -npts, and also tried multiplying the audio signal, [*~] , before going into [sigmund~] but I notices no difference in pitch response, although I noticed that the envelope outlet had a raised value, I made sure the mac laptop mic had noise reduction off.

I was hoping to hear layers and clusters of sinus tones as [sigmund~] would try to track the noise from raised gain microphone signal.

So I have searched around the forum and downloaded a few patches but I have not been able to figure out how to set up [sigmund~] to achieve something along the lines of what I am trying to do.

Help plz?

@cfry Laptop mics are meant for nearfield pickup of people/things near the screen and are not sensitive enough to pickup much from any distance, so you increase the gain which also increases background noise and [sigmund~] can not differentiate between the background noise and birdsong any longer. Turning noise reduction back on may help, it might be smart enough to lock in on the birdsong, but a more sensitive microphone and a interface with decent noise specs will do you better.

@oid said:

@cfry ... [sigmund~] can not differentiate between the background noise and birdsong any longer.
Ok, but I really would like polyphonic tracking of all/any background noise. Maybe [sigmund~] is not the tool for this.

@cfry The background noise includes things like fan and hard drive noises from your computer, white noise that is generated by the mic and its amplifier, and wind noise, if all of these noises are at or near equal in volume to the birdsong, [sigmund~] can not pick them out, even if you can hear them through the noise, our brains are good at filtering out what we do not want to hear. A quick test shows my cheap MXL large diaphragm condenser mic and Behringer interface is sensitive and quiet enough for [sigmund~] to get the crow that is cawwing away right now even with my windows closed, would pickup the quieter birds no problem as well if the windows were open and the wind/waves were not up. [sigmund~] will do the job, but it's output is only as good as its input.

@oid Thanks a lot, really good to know that [sigmund~] works that good with decent audio equipment.

But I am not after bird song, I actually want to track the wind, cars passing by, all sounds in a sound field. Of course there are no clear pitches to analyze and synthesize in those type of sounds, but I was hoping for a reaction from the patch, for a start. I think that TimbreID would let me do what I hope for.

Still it is great to know that there is such a difference with the right gear.

@cfry Those sorts of sounds can be gotten as well and a more sensitive mic will pick them up. The problem you are having is the mic is just getting noise, everything of the same volume, so nothing distinct for [sigmund~] to pick out, remember all those discrete sources add together. By wind noise I meant the sound of wind hitting a microphone directly, this creates a constant sound which will over power all those other sounds, not the sound of wind rustling the leaves in a tree. A good wind screen will be very helpful for you here, but you need to remember, a windscreen does reduce mic sensitivity, so there is a trade off, increase mic sensitivity and the more wind noise it picks up, put on a denser wind screen and you loose some of that sensitivity. A pop screen could work better since it can be placed between the wind and the mic, the other sides of the mic are left open, but if the wind shifts you could end up with wind noise overpowering everything and have to reposition the screen. If you limit yourself to days with nothing more than light winds, you should be able to get by with just a light windscreen and not suffer much loss in sensitivity.

My knowledge of windscreens and pop screens and the like is fairly limited and largely theoretical, I have little hands on use of these things as I mostly record in more controlled environments or in situations where I have more leeway than your needs allow. Seeking out people or sites dedicated to making field recordings would likely be your best path on finding a good mic/pre setup for your needs.

One thing that could work would be to do a fft, and look for patterns and similarities in the amplitude enveloppe. A trained neural network could do so. I've read somewhere that they use them to analyse highly noisy signal like satellite transmissions.

@Il-pleut TimbreID http://williambrent.conflations.com/pagez/research.html#timbreID
works along those lines, not machine learning but it compares features of sounds, and do real time concatenative synthesis and timbre-based orderings of sound sets among other things. And it works really well I say out of my initial fooling around with it. Thanks @Johnny-Mauser for letting me know it existed. If TimbreID could be used to both analyze and process in realtime I am still to find out. Maybe that is where a trained neural network is needed.

Seems cool.