[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: SNR estimation from noisy data

Hi Tarun
I assume you are referring to a speech signal in a somewhat more stationary background noise? In this case, the general idea is to estimate the noise level from the long term average level of the speech + noise mixture. Then short term fluctuations from this average are the signal component. Thus you can estimate the SNR.
For a good algorithm, I would start with this paper:
Title: Noise estimation by minima controlled recursive averaging for robust speech enhancement
Author(s): Cohen I, Berdugo B
Source: IEEE SIGNAL PROCESSING LETTERS   Volume: 9   Issue: 1   Pages: 12-15   Published: JAN 2002
There are also lots of noise estimators you might find in the hearing aid literature.
The difficulty occurs when the signal and noise become similar, such as a competing voice. In fact, consider the unfortunate cocktail party situation where a single competing talker ("background noise") might be saying something more interesting than the actual target speaker, and so the SNR spontaneously inverts. There is little chance of estimating SNR in this situation.

From: AUDITORY - Research in Auditory Perception [mailto:AUDITORY@xxxxxxxxxxxxxxx] On Behalf Of Tarun Pruthi
Sent: Wednesday, 6 May 2009 12:44 AM
To: AUDITORY@xxxxxxxxxxxxxxx
Subject: [AUDITORY] SNR estimation from noisy data

Hi all:

Could anyone point me to the best algorithms available to estimate SNR from noisy data?

Senior Research Engineer
Think A Move, Ltd.