From: bharat pathak on
I am trying to isolate the silence regions from the speech
regions. For this I am trying to find the enevelope of speech
signal and compare it with set threshold value. My question is
what frequencies does a speech envelope typically contain?

What are the typical ways in which people perform the isolation
of speech signals from the silence regions. Assume in some
presence of noise.

bharat
From: Jerry Avins on
On 3/31/2010 10:29 PM, bharat pathak wrote:
> I am trying to isolate the silence regions from the speech
> regions. For this I am trying to find the enevelope of speech
> signal and compare it with set threshold value. My question is
> what frequencies does a speech envelope typically contain?
>
> What are the typical ways in which people perform the isolation
> of speech signals from the silence regions. Assume in some
> presence of noise.
>
> bharat

What do you mean by the envelope of a baseband signal? That's a new one
on me.

Jerry
--
"It does me no injury for my neighbor to say there are 20 gods, or no
God. It neither picks my pocket nor breaks my leg."
Thomas Jefferson to the Virginia House of Delegates in 1776.
���������������������������������������������������������������������
From: Rafael Deliano on
> I am trying to isolate the silence regions from the speech
> regions.

Typical applications are:
a) old speech transmission systems ( TASI ) ;
speakerphones ; in modern speech transmission more
often called VAD
http://en.wikipedia.org/wiki/Voice_activity_detection
b) isolated word speech recognition.

> the enevelope of speech
> signal and compare it with set threshold value.

Energy based detection is usable for a) because the human
listener will not mind too much about missing short
low-energy/high-frequency parts.
For b) two signals like energy + zero-crossing-rate
are preferred.

> Assume in some presence of noise.
Makes energy less preferable. If you define the noise as
low-frequency ( Hoth-noise ) then zero-crossing would
still be usable.

MfG JRD
From: HardySpicer on
On Apr 1, 3:29 pm, "bharat pathak" <bharat(a)n_o_s_p_a_m.arithos.com>
wrote:
> I am trying to isolate the silence regions from the speech
> regions. For this I am trying to find the enevelope of speech
> signal and compare it with set threshold value. My question is
> what frequencies does a speech envelope typically contain?
>
> What are the typical ways in which people perform the isolation
> of speech signals from the silence regions. Assume in some
> presence of noise.
>
> bharat

Hige literature on this one. Voice-activity detectors - look on Google
Scholar.


Hardy
From: Clay on
On Mar 31, 10:29 pm, "bharat pathak" <bharat(a)n_o_s_p_a_m.arithos.com>
wrote:
> I am trying to isolate the silence regions from the speech
> regions. For this I am trying to find the enevelope of speech
> signal and compare it with set threshold value. My question is
> what frequencies does a speech envelope typically contain?
>
> What are the typical ways in which people perform the isolation
> of speech signals from the silence regions. Assume in some
> presence of noise.
>
> bharat

POTS (Plain Old Telephone Service) passes just below 300 to just above
3200 Hz, so expect most of your speech to be in that frequency range.

Clay