Searching \ for '[OT] Noise and spike removal in audio, and the dec' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/io/audio.htm?key=audio
Search entire site for: 'Noise and spike removal in audio, and the dec'.

Exact match. Not showing close matches.
PICList Thread
'[OT] Noise and spike removal in audio, and the dec'
2010\11\30@002901 by V G

picon face
Hi all,

I've been recording lectures for my classes using the Zoom H2 (
http://www.zoom.co.jp/english/products/h2/) portable audio recorder. I
normally sit near the front (no more than 5 or 6 m away from the professor)
and therefore tend to catch a lot of background noise, as well as noise from
other students.

1. At this distance, the high gain mode is more than enough to capture what
the professor is saying, however if someone sitting near the recorder coughs
or sneezes or flips a page or smashes their face against their desk, I get
high amplitude clicks in the recording. Such noises are far louder than the
sound from the professor. Because of this, I find it hard to "normalize" the
recording using Audacity since the maximum amplitude wont be adjusted to
more than 0 db.

How would i clip off these loud spikes? In Audacity, they're clearly visible
as tall spikes, compared to the rest of the recording.

2. Audacity also has a really nice noise removal function. You select a
section of audio (as a sample) that is "noise", and it filters that noise
out from the entire recording. For lectures, the beginning and end of the
lectures are noisy, as everyone settles down. The quietest sections can be
found in the middle of the lectures. However, during that time, I find it
hard to find a gap where it's *only* noise and the professor isn't talking
(since he is almost always speaking). Any ideas as to how to remove this
noise?

3. I have a question about decibels. I understand that when measuring volume
or sound intensity, 0 dB is the level at which sound is barely audible for
the average human. So, any sound that is louder will use that barely audibly
sound as a reference on the decibel scale.

3.1. However, the normalizing function says it will stretch the maximum
amplitude to 0 dB. What does this mean? What is the reference?

3.2. My recorder has a sound intensity scale (I think) that is showing
values from -48 dB all the way up to 0 dB on the high end. You can see the
bar moving higher if you speak louder. I don't understand this scale

2010\11\30@022423 by Richard Prosser

picon face
VG

I'll try & answer some of your questions.

On 30 November 2010 17:28, V G <spam_OUTx.solarwind.xTakeThisOuTspamgmail.com> wrote:
{Quote hidden}

..Some sort of directional mic (preferably with built-in noise
cancelling) will help getting the desired audio while minimising the
background. It may not be practical in your situation but would give a
massive improvement.


>
> 2. Audacity also has a really nice noise removal function. You select a
> section of audio (as a sample) that is "noise", and it filters that noise
> out from the entire recording. For lectures, the beginning and end of the
> lectures are noisy, as everyone settles down. The quietest sections can be
> found in the middle of the lectures. However, during that time, I find it
> hard to find a gap where it's *only* noise and the professor isn't talking
> (since he is almost always speaking). Any ideas as to how to remove this
> noise?
>
Audacity lets you amplify beyond clipping. There is a tick box and you
can amplify as required. You can bank out the noise peaks before
amplifying or you can reduce the harshness by low pass filtering the
clipped result after amplifiction. Experiment to find the best cutoff
frequency & rolloff rate.

{Quote hidden}

0dB is simply a reference point. There is a tendency to use it as a
sound pressure level of 1pascal/m^2 or something close to the audible
threshold, but unless the reference is quoted it has no absolute
value. Audacity uses the clipping level as the 0dB reference point and
levels measure downward from here (negative values).

So Audacity is showing sound levels from 0 to -48dB or so and others
are using the threshold of hearing as 0dB and it goes up from there -
e.g to a jet engine of +140dB or whatever.

To emphasize, dBs are a measure of power ratio - not absolute levels.
In electronics power may be expressed as dBm or dBV etc. Here the
reference level IS included as the (m)illiwatt or (V)olt etc.

Audio levels can also be expressed as dBA or dBC etc. In this case the
reference level IS standardised, along with the frequency response of
the listening device. I'll leave it up to you to google the details.

Hope this assists.

R

2010\11\30@031941 by Michael Watterson

face picon face
 On 30/11/2010 07:23, Richard Prosser wrote:
> VG
>
> I'll try&  answer some of your questions.
>
> O
>> kes, compared to the rest of the recording.
> .Some sort of directional mic (preferably with built-in noise
> cancelling) will help getting the desired audio while minimising the
> background. It may not be practical in your situation but would give a
> massive improvement.
>
>
Noise cancelling mics are for close to the mouth. no use here

Yes. A directional mic.
The more directional, the larger it is and the more accurately it needs pointed. A cardioid is similar size to an ordinary microphone.
Stereo or Binaural (better) recording helps as then on listening the "cocktail" effect works in presence of noise

2010\11\30@044407 by V G

picon face
On Tue, Nov 30, 2010 at 3:19 AM, Michael Watterson <.....mikeKILLspamspam@spam@radioway.org>wrote:

> Yes. A directional mic.
> The more directional, the larger it is and the more accurately it needs
> pointed. A cardioid is similar size to an ordinary microphone.
> Stereo or Binaural (better) recording helps as then on listening the
> "cocktail" effect works in presence of noise


I'm not going to get another microphone. The microphone that I have now is
exceptionally good. I just want to know how I can improve the quality of the
sound after it has been recorded

2010\11\30@044501 by alan.b.pearce

face picon face
> 3. I have a question about decibels. I understand that when measuring
volume
> or sound intensity, 0 dB is the level at which sound is barely audible
for
> the average human. So, any sound that is louder will use that barely
audibly
> sound as a reference on the decibel scale.
>
> 3.1. However, the normalizing function says it will stretch the
maximum
> amplitude to 0 dB. What does this mean? What is the reference?
>
> 3.2. My recorder has a sound intensity scale (I think) that is showing
> values from -48 dB all the way up to 0 dB on the high end. You can see
the
> bar moving higher if you speak louder. I don't understand this scale.

Answering 3.2 first ...

The designers of your recorder will have ascertained a maximum level at
which to record. I assume this recorder is using tape as the storage
medium, and will answer on that basis, but if the storage medium is some
form of digital signal a similar thing applies. With tape, because the
storage is a magnetic medium it is possible to increase the recording
level, but instead of the signal hard clipping at a maximum level you
get what is sometimes referred to as 'soft clipping', where the peak of
the waveform is at a lower level than it should be, but is still rounded
rather than a flat top. The effect to the ear is a muddying of the
signal, and if checking the distortion with a meter, the amount of
distortion rises rapidly for a small increase in signal level. What the
designer will do is determine a maximum level of distortion that is
permitted, ascertain the signal level required for this, and then set
the 0dB recording level around 3dB below this, so there is a wee bit of
head room available for peak signals in speech and the like. The
recording level meter is then calibrated with this 0dB near maximum
deflection (or nearly all leds lit, as you seem to be using a LED level
meter) and then calibrated in -dB markings from there, so you can see
how much you can increase the signal input level to get best signal to
noise when listening to the recording.

Now to answer 3.1 ...

I don't know for sure, but it sounds like Audacity is treating the 0dB
level the same way as the recording level meter, i.e. 0dB is the level
at which the digital signal will still produce a waveform without
clipping. So if the digital signal has an amplitude that is well below
the maximum that the DAC can handle, then the normalizing function in
Audacity will multiply all the samples by some factor that will make the
maximum amplitude ones reach full scale for the DAC, thereby making the
signal as loud as possible.
-- Scanned by iCritical.

2010\11\30@050044 by William \Chops\ Westfield

face picon face

On Nov 30, 2010, at 12:19 AM, Michael Watterson wrote:

> A directional mic.

So does anyone make small, moderately directional mics aimed at  recording this sort of thing (spoken material, not music..)  Most of  what I've seen has cheap to not-so-cheap omnidirectional mics, leading  up to more specific mics aimed at professional music recording and  such (one "review of low-cost shotgun mics" had in mind a range of  "$250 - $400" for "low cost."  Ouch.)  Most things I've seen seem to  work by physically filtering noise from the edges and using a very  sensitive mic element ("we're restricting our input to 1/100 of the  sphere of an omnidirectional mic, so we need to be 100 times more  sensitive overall" ?)  I'd think that a modern DSP with an array of  moderately sensitive omnidirectional mics ought to be able to do  interesting things with phase ("noise canceling", but for a different  definition of "noise" than usually used...)

BillW

2010\11\30@054446 by RussellMc

face picon face
>> Yes. A directional mic.
>> The more directional, the larger it is and the more accurately it needs
>> pointed. A cardioid is similar size to an ordinary microphone.
>> Stereo or Binaural (better) recording helps as then on listening the
>> "cocktail" effect works in presence of noise

> I'm not going to get another microphone. The microphone that I have now is
> exceptionally good. I just want to know how I can improve the quality of the
> sound after it has been recorded.

You may be choosing between prettiness/niceness/"goodnes" and functionality..

BUT a much lower "goodness" low cost microphone may do a better job
for your purpose.

The sound in the room has spatially dependent qualities (direction of
arrival, path taken,...) and much of that is lost at the moment it is
recorded by even the world's best mono microphone. Using a stereo
microphone (which yours may be) may well help and a noise cancelling
microphone (where relevant) actually does "processing" outside the
recorder.

A potential means of improvement is to add some form of reflector
9gain-antenna effectively) which both reduces off subject noise and
boosts desired signal. You may need to wear a tinfoil hat to distract
attention from the reflector if you use a large one.

Build your own.
WARNING - DO NOT TRY TO CARRY THIS ONTO AN AIRCRAFT.
Your prof may not be impressed.

   http://blockyourid.com/~gbpprorg/mil/shotgun/index.html


Nor with this :-).
But, a smallish version may have some advantage

 http://upload.wikimedia.org/wikipedia/commons/3/3e/ParabolicMicrophone.jpg

2010\11\30@090859 by Olin Lathrop

face picon face
V G wrote:
> I'm not going to get another microphone. The microphone that I have
> now is exceptionally good.

Not really.  Not for this purpose anyway.  "Good" for a microphone usually
means flat frequency response from 20Hz to 20KHz.  That kind of performance
doesn't help much in this case.  Directionality is a much better measure of
"good" for your particular problem.

> I just want to know how I can improve the
> quality of the sound after it has been recorded.

As stated before, that's much harder and success will be limited.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

2010\11\30@091447 by Lee Jones

flavicon
face
>> Yes. A directional mic.
>
> I'm not going to get another microphone. The microphone that I
> have now is exceptionally good.

Apparently (from your comments) excellent at picking up sounds
accurately from all directions.  What is considered excellent in
one application may not be the best tool for another application.

> I just want to know how I can improve the quality of the sound
> after it has been recorded.

Quick answer is -- you can't.  That's an over simplification.
You can but it's difficult, time consuming, and/or expensive --
sometimes all 3 of the above.

Using a photography analogy, you can use a great camera to take
a poorly lit, unfocused image.  Then spend vast amounts of time
in Photoshop to "fix" the lighting & use deconvolution plug-ins
to fix the focus.  Once all that work is done, you'll have a
photo that is almost as good as if you had just properly lit
the scene (or used the right exposure) and focused the camera.
I'm of the camp that believes in getting it right at the
moment the shutter is clicked.  It saves lots of headaches.

You might be able to get way more "bang for your buck" by
switching to a directional microphone for future recordings.

                                               Lee Jone

2010\11\30@092522 by alan.b.pearce

face picon face
{Quote hidden}

I cannot help but get the feeling that the recording is being
surreptitiously, and changing to a better microphone, which will then be
in an obvious position, will make it obvious what is being done ... ;))
-- Scanned by iCritical.

2010\11\30@203937 by RussellMc

face picon face
BCCs: VMP Ki (Ro) - St Andrews

> Besides, it wouldn't  take experts on the other side long to figure out what the funny looking
> large hexagonal panels high on the sides of the ship are for.

Targets, I believe.
(The other guys seem to think so, anyway.)

For a VERY large distributed version see the VLBA

        http://en.wikipedia.org/wiki/Very_Long_Baseline_Array

Sending one Aegis Frigate "as a sign of support" and/or on a goodwill
tour and/or "to monitor activitiies"  is seen as sabre rattling and
war mongering.
Sending two is seen as close to mobilisation and not too far below a
declaration of war.
Sending three - I don't think they do.
One Aegis Frigate can (has) remove(d) a satellite from LEO using a
"Standard missile".
Two Aegis frigates ... :-) :-(.

Long ago a Frigate was a feared and dangerous fighting machine in the
right context*.
Latterly it became a poor man's destroyer.
Things seem to have gone full circle


        Russell

* These people thought so:

3 British Frigates:

Pretty pix:  http://www.catswhiskerstours.co.uk/tours_subdyn.asp?inst=4&id=3194

www.infobritain.co.uk/Eilean_Donan_Castle.htm
www.electricscotland.com/historic/castles/eilean_donan.htm
http://www.aboutaberdeen.com/eileandonancastle.php

French & British Frigates
~463 years ago

http://www.jstor.org/pss/576909
http://ehr.oxfordjournals.org/content/CXI/442/578.extract

The subsequent founder of the Presbyterian church was on the wrong end
of the frigates' guns during this occasion.







'[OT] Noise and spike removal in audio, and the dec'
2010\12\01@075301 by Olin Lathrop
face picon face
RussellMc wrote:
> Sending one Aegis Frigate "as a sign of support" and/or on a goodwill
> tour and/or "to monitor activitiies"  is seen as sabre rattling and
> war mongering.
> Sending two is seen as close to mobilisation and not too far below a
> declaration of war.

Which is most likely exactly what is intended.  "Rethink what you're doing,
or else...".

This sort of tactic has been used many times thru history by many nations,
and is hardly limited to Arleigh Burke class destroyers.  Sometimes it
works.


********************************************************************
Embed Inc, Littleton Massachusetts, http://www.embedinc.com/products
(978) 742-9014.  Gold level PIC consultants since 2000

More... (looser matching)
- Last day of these posts
- In 2010 , 2011 only
- Today
- New search...