Outdoor recording question

[ric29]

I have got a recording which has been done during windy day so the sound of the wind sometimes covers people speech.
What is the best method or the good point to start to decrease sound of the wind?
Thanx

[Craig Maier]

Hi,

I think that the best approach to a noise problem like that is the use of the AFDF (Adaptive Frequency Domain Filter) found under the Continuous Noise Filter in the Forensics version of the software or the Time Domain Adaptive Filter found under the Forensics Menu in the Forensics Version or the Auto-Voice filter found under the Forensics Menu in the Forensics version. If you do not have a copy of the Forensics version, you can download a demo version from our website located at www.diamondcut.com. It will work for around 10 days as the demo.

I hope that helps,

Craig

[Dan McDonald]

and this will sound odd but you can sometimes reduce windnoise by converting the lower frequencies to mono using the channel blender. I have no idea why that works, but I've managed to reduce windnoise a lot in doing that. I found it when I was trying to reduce mic handling noise with hand-held mics.

[Marc Hildebrant]

Dan,

When you say "lower frequencies", do you mean the speech frequencies ?

Marc

[Dan McDonald]

I put it about 150-250 Hz and lower, so lower than most speech I think.

[Craig Maier]

Dan,

Doesn't that method using the Channel Blender only work with a stereo or binaural signal?

Craig

[Craig Maier]

Hi Again,

I suppose that a steep high pass filter (24 dB/Oct, Chebyshev) set for around 150 Hz may help a bit too.

Craig

[Dan McDonald]

It seems to work with a mono file transferred to stereo. I am not sure why. I've only tried it with cassette transfers, so maybe only if there's an azimuth issue? Would that make sense?

[Craig Maier]

I think that something would have to create a phase shift between the signals on each channel for low frequency rejection to occur using the Channel Blender. If the signal was pure mono with no phase shift, I think that the system would not sum to mono with a reduction effect of the lower frequencies. I guess that a Cassette tape that is off-azimuth could produce enough phase shift to provide a benefit, however. But, I think it would take a lot of phase shift and thus a real bad azimuth situation.

I use that channel blender feature mostly to reduce record rumble noise since turntable and record cutter rumble are mostly in the vertical plane. So, summing to mono with a mechanical record recording causes the vertical component to be reduced by cancellation (since the two channels comming from the record cartridge are out-of-phase with respect to the vertical plane of the record and in-phase with respect to lateral motion). But, I have heard of the use of the Channel Blender as described when two microphones are used into a stereo recorder on a windy day. I guess it depends on the relative positioning of the mics as to whether or not this wind noise reduction method will work in that situation. Of course, if the wind noise is completely random with respect to both mics, then cancellation can not occur.

Craig

[Craig Maier]

Hi Again,

As for speech frequencies, there is something in the helpfile about that, I think in the glossary section - - - maybe under formants. As I recall, the lowest speech frequency is roughly 90 Hz with the extreme end of the spectrum of a mature male voice.

Craig

[Dan McDonald]

Yes, I have had situations where it didn't help at all, and some situations where it made a big difference. I haven't investigated to figure out the 'why' part of it.

[Craig Maier]

Hi,

I just found the section that describes speech frequencies in terms of the lowest values. It was under Voice Fundamental Frequencies in the glossary:



Voice Fundamental Frequency (F0)

The fundamental frequency of a human voice is sometimes referred to as F0. For adult females, that range runs from 165 to 255 Hz and for adult males the range runs from 85 to 180 Hz. Speech articulation uniqueness is the result of the ratio of formant frequencies F1 through F4.

[ric29]

Hi,

I think that the best approach to a noise problem like that is the use of the AFDF (Adaptive Frequency Domain Filter) found under the Continuous Noise Filter in the Forensics version of the software or the Time Domain Adaptive Filter found under the Forensics Menu in the Forensics Version or the Auto-Voice filter found under the Forensics Menu in the Forensics version. If you do not have a copy of the Forensics version, you can download a demo version from our website located at www.diamondcut.com. It will work for around 10 days as the demo.

Thanx a lot Craig.
You are always the great source of knowledge
I'll do it in this way.

[ric29]

and this will sound odd but you can sometimes reduce windnoise by converting the lower frequencies to mono using the channel blender. I have no idea why that works, but I've managed to reduce windnoise a lot in doing that. I found it when I was trying to reduce mic handling noise with hand-held mics.

Although my recorded signal is mono so can't be used in my case, this method should be definitely remembered just in case.
Thanx Dan.

[Craig Maier]

The new DCForensics10.xx version of the software has a special Dynamic Wind Noise Filter which only invokes itself in the presence of wind noise, preserving the bottom end of the audio spectrum as best as possible.

Craig