| View previous topic :: View next topic |
| Author |
Message |
SRoe
Guest
|
 Posted: Sun Nov 16, 2008 10:13 pm Post subject: Noisy or quiet environment |
 |
|
Hi,
I am trying to write a piece of software which analyses a recording of a
conversation and decides whether the conversation was in a noisy
environment (eg next to a busy road) or whether the conversation was in
a quiet environment (eg an empty room).
I have test recordings for each type which I made using my PDA with an
8kHz 16bit mono format.
I'm a newbie to DSP and don't have a clue where to start or even where
to begin to look. I'm unsure to what charactertics I should be looking
for. Any help and pointers would be highly appreciated.
--
Many thanks,
Stephen |
|
| |
|
 Back to top |
prathamesh
Guest
|
| Posted: Sun Nov 16, 2008 10:13 pm Post subject: Re: Noisy or quiet environment |
|
|
On Nov 16, 9:13 pm, SRoe <s...@example.net> wrote:
| Quote: | Hi,
I am trying to write a piece of software which analyses a recording of a
conversation and decides whether the conversation was in a noisy
environment (eg next to a busy road) or whether the conversation was in
a quiet environment (eg an empty room).
I have test recordings for each type which I made using my PDA with an
8kHz 16bit mono format.
I'm a newbie to DSP and don't have a clue where to start or even where
to begin to look. I'm unsure to what charactertics I should be looking
for. Any help and pointers would be highly appreciated.
--
Many thanks,
Stephen
|
If you are going to be specific about the environment in which you are
going to test your software(i.e in this case if it is always going to
be on busy road) you can do the frequency analysis of the just the
noise signal by recording only the noise occurring on a busy
road ......this will provide you the frequency range in which the
noise signals lie and will help you in predicting the kind of
conversation..... |
|
| |
|
| Back to top |
Jerry Avins
Guest
|
| Posted: Mon Nov 17, 2008 12:24 am Post subject: Re: Noisy or quiet environment |
|
|
SRoe wrote:
| Quote: | Hi,
I am trying to write a piece of software which analyses a recording of a
conversation and decides whether the conversation was in a noisy
environment (eg next to a busy road) or whether the conversation was in
a quiet environment (eg an empty room).
I have test recordings for each type which I made using my PDA with an
8kHz 16bit mono format.
I'm a newbie to DSP and don't have a clue where to start or even where
to begin to look. I'm unsure to what charactertics I should be looking
for. Any help and pointers would be highly appreciated.
|
What does "quiet" mean? Can you distinguish the quiet background from
the noisy by looking for intervals of silence? speech is rarely
continuous for long.
Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ |
|
| |
|
| Back to top |
Tim Wescott
Guest
|
| Posted: Mon Nov 17, 2008 1:11 am Post subject: Re: Noisy or quiet environment |
|
|
On Sun, 16 Nov 2008 16:13:52 +0000, SRoe wrote:
| Quote: | Hi,
I am trying to write a piece of software which analyses a recording of a
conversation and decides whether the conversation was in a noisy
environment (eg next to a busy road) or whether the conversation was in
a quiet environment (eg an empty room).
I have test recordings for each type which I made using my PDA with an
8kHz 16bit mono format.
I'm a newbie to DSP and don't have a clue where to start or even where
to begin to look. I'm unsure to what charactertics I should be looking
for. Any help and pointers would be highly appreciated.
|
Have you done a web search?
In general when you want to distinguish one signal from another you want
to figure out how the signals are similar (so you can discount the
similarities), and how the signals are different (so you can look for the
differences).
Jerry pointed out a very good characteristic of speech, which is that it
is not continuous -- so you can distinguish between the part of the
signal that is episodic (the speech) from the part of the signal that is
continuous (the noise, we hope).
I would also suggest that all the voiced parts of speech (i.e. anything
that uses the vocal cords) tends to have a rich harmonic content, because
your vocal cords are basically a cheap buzzer, whose sound gets filtered
by your vocal tract. So an FFT of voice will contain a lot of spikes,
while noise may or may not. (The difficulty here is that voice also has
sibilants, which you make with your lips and tongue, and which have a
pretty white spectrum).
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html |
|
| |
|
| Back to top |
itakatz
Guest
|
| Posted: Mon Nov 17, 2008 1:52 pm Post subject: Re: Noisy or quiet environment |
|
|
On Nov 16, 9:11 pm, Tim Wescott <t...@justseemywebsite.com> wrote:
| Quote: | On Sun, 16 Nov 2008 16:13:52 +0000, SRoe wrote:
Hi,
I am trying to write a piece of software which analyses a recording of a
conversation and decides whether the conversation was in a noisy
environment (eg next to a busy road) or whether the conversation was in
a quiet environment (eg an empty room).
I have test recordings for each type which I made using my PDA with an
8kHz 16bit mono format.
I'm a newbie to DSP and don't have a clue where to start or even where
to begin to look. I'm unsure to what charactertics I should be looking
for. Any help and pointers would be highly appreciated.
Have you done a web search?
In general when you want to distinguish one signal from another you want
to figure out how the signals are similar (so you can discount the
similarities), and how the signals are different (so you can look for the
differences).
Jerry pointed out a very good characteristic of speech, which is that it
is not continuous -- so you can distinguish between the part of the
signal that is episodic (the speech) from the part of the signal that is
continuous (the noise, we hope).
I would also suggest that all the voiced parts of speech (i.e. anything
that uses the vocal cords) tends to have a rich harmonic content, because
your vocal cords are basically a cheap buzzer, whose sound gets filtered
by your vocal tract. So an FFT of voice will contain a lot of spikes,
while noise may or may not. (The difficulty here is that voice also has
sibilants, which you make with your lips and tongue, and which have a
pretty white spectrum).
--
Tim Wescott
Wescott Design Serviceshttp://www.wescottdesign.com
Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says..
See details athttp://www.wescottdesign.com/actfes/actfes.html
|
Hi,
If the level of the speech signal is more or less constant between
different experiments, you can simply calculate the signal-to-noise
ratio, for example by averaging the energy (signal squared) over a
finite frame length, and taking the ratio between the 'loudest' frame
and the 'quietest' frame. |
|
| |
|
| Back to top |
jim
Guest
|
| Posted: Mon Nov 17, 2008 7:28 pm Post subject: Re: Noisy or quiet environment |
|
|
Hi Stephen,
Speech derives much of it's meaning from periods of silence. Besides just
pauses for emphasis and separation of words or syllables there are quite a few
common sounds like B, P, M, T, D, N, K, that are produced by briefly
interrupting the flow of air and thus produce some few milliseconds of quiet.
In an 8 kHz sample rate recording of speech with no background noise you should
see at least several episodes a second where the amplitude drops to less than
some minimum threshold (maybe 2%-5% of the mean) for some short period (maybe
20-80 contiguous samples). Of course noise might have interruptions also, but
usually not so common or consistent.
-jim
SRoe wrote:
| Quote: |
Hi,
I am trying to write a piece of software which analyses a recording of a
conversation and decides whether the conversation was in a noisy
environment (eg next to a busy road) or whether the conversation was in
a quiet environment (eg an empty room).
I have test recordings for each type which I made using my PDA with an
8kHz 16bit mono format.
I'm a newbie to DSP and don't have a clue where to start or even where
to begin to look. I'm unsure to what charactertics I should be looking
for. Any help and pointers would be highly appreciated.
--
Many thanks,
Stephen
|
----== Posted via Pronews.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.pronews.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= - Total Privacy via Encryption =--- |
|
| |
|
| Back to top |
|
