Subject: Re: SNR estimation from noisy data From: Daniel TAFT <DTAFT@xxxxxxxx> Date: Wed, 6 May 2009 10:32:27 +1000 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>This is a multi-part message in MIME format. ------_=_NextPart_001_01C9CDE2.221F986E Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Hi Tarun =20 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. =20 For a good algorithm, I would start with this paper: =20 Title: Noise estimation by minima controlled recursive averaging for robust speech enhancement <http://apps.isiknowledge.com.ezp.lib.unimelb.edu.au/full_record.do?prod uct=3DWOS&search_mode=3DGeneralSearch&qid=3D1&SID=3D1DcacECO7BBeLiPcIED&p= age=3D1&d oc=3D4> =20 Author(s): Cohen I, Berdugo B=20 Source: IEEE SIGNAL PROCESSING LETTERS Volume: 9 Issue: 1 Pages: 12-15 Published: JAN 2002=20 =20 There are also lots of noise estimators you might find in the hearing aid literature. =20 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. =20 Regards, Daniel. =20 =20 ________________________________ From: AUDITORY - Research in Auditory Perception [mailto:AUDITORY@xxxxxxxx On Behalf Of Tarun Pruthi Sent: Wednesday, 6 May 2009 12:44 AM To: AUDITORY@xxxxxxxx Subject: [AUDITORY] SNR estimation from noisy data Hi all:=20 Could anyone point me to the best algorithms available to estimate SNR from noisy data? Thanks Tarun Senior Research Engineer Think A Move, Ltd. ------_=_NextPart_001_01C9CDE2.221F986E Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN"> <HTML><HEAD> <META http-equiv=3DContent-Type content=3D"text/html; = charset=3Dus-ascii"> <META content=3D"MSHTML 6.00.6000.16809" name=3DGENERATOR></HEAD> <BODY> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><FONT = face=3DArial=20 color=3D#0000ff size=3D2>Hi Tarun</FONT></SPAN></DIV> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><FONT = face=3DArial=20 color=3D#0000ff size=3D2></FONT></SPAN> </DIV> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><FONT = face=3DArial=20 color=3D#0000ff size=3D2>I assume you are referring to a speech signal = in a somewhat=20 more stationary background noise? </FONT></SPAN><SPAN=20 class=3D389151700-06052009><FONT face=3DArial color=3D#0000ff = size=3D2>In this case, the=20 general idea is to estimate the noise level from the long term average = level of=20 the speech + noise mixture. Then short term fluctuations from this = average are=20 the signal component. Thus you can estimate the SNR.</FONT></SPAN></DIV> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><FONT = face=3DArial=20 color=3D#0000ff size=3D2></FONT></SPAN> </DIV> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><FONT = face=3DArial=20 color=3D#0000ff size=3D2>For a good algorithm, I would start with this=20 paper:</FONT></SPAN></DIV> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><FONT = face=3DArial=20 color=3D#0000ff size=3D2></FONT></SPAN> </DIV> <DIV dir=3Dltr align=3Dleft><SPAN class=3D389151700-06052009><SPAN = id=3Drecords_chunks=20 style=3D"DISPLAY: block">Title: <A class=3DsmallV110=20 oncontextmenu=3D"javascript:return IsAllowedRightClick(this);"=20 href=3D"http://apps.isiknowledge.com.ezp.lib.unimelb.edu.au/full_record.d= o?product=3DWOS&search_mode=3DGeneralSearch&qid=3D1&SID=3D1Dc= acECO7BBeLiPcIED&page=3D1&doc=3D4"><FONT=20 color=3D#000000>Noise estimation by minima controlled recursive = averaging for=20 robust speech enhancement</FONT></A> <BR>Author(s): <SPAN = class=3DhitHilite>Cohen=20 I</SPAN>, <SPAN class=3DhitHilite>Berdugo B</SPAN> <BR>Source: <SPAN=20 class=3Ddata_bold>IEEE SIGNAL PROCESSING LETTERS</SPAN> = Volume: <SPAN=20 class=3Ddata_bold>9</SPAN> Issue: <SPAN = class=3Ddata_bold>1</SPAN>=20 Pages: <SPAN class=3Ddata_bold>12-15</SPAN> = Published:=20 <SPAN class=3Ddata_bold>JAN 2002</SPAN> </SPAN></SPAN></DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2></FONT> </DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN = class=3D389151700-06052009>There=20 are also lots of noise estimators you might find in the hearing aid = literature.</SPAN></FONT></DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20 class=3D389151700-06052009></SPAN></FONT> </DIV> <DIV><FONT><SPAN class=3D389151700-06052009></SPAN></FONT><FONT = face=3DArial=20 color=3D#0000ff size=3D2><SPAN class=3D389151700-06052009>The difficulty = occurs when=20 the signal and noise become similar, such as a competing voice. In fact, = consider the unfortunate cocktail party situation where a single = competing=20 talker ("background noise") might be saying something more = interesting than=20 the actual target speaker, and so the SNR spontaneously inverts. There = is little=20 chance of estimating SNR in this situation.</SPAN></FONT></DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20 class=3D389151700-06052009></SPAN></FONT> </DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20 class=3D389151700-06052009>Regards,</SPAN></FONT></DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20 class=3D389151700-06052009>Daniel.</SPAN></FONT></DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20 class=3D389151700-06052009></SPAN></FONT> </DIV> <DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20 class=3D389151700-06052009></SPAN></FONT> </DIV> <DIV><BR></DIV> <DIV class=3DOutlookMessageHeader lang=3Den-us dir=3Dltr align=3Dleft> <HR tabIndex=3D-1> <FONT face=3DTahoma size=3D2><B>From:</B> AUDITORY - Research in = Auditory Perception=20 [mailto:AUDITORY@xxxxxxxx <B>On Behalf Of </B>Tarun=20 Pruthi<BR><B>Sent:</B> Wednesday, 6 May 2009 12:44 AM<BR><B>To:</B>=20 AUDITORY@xxxxxxxx<BR><B>Subject:</B> [AUDITORY] SNR estimation = from noisy=20 data<BR></FONT><BR></DIV> <DIV></DIV><SPAN class=3DApple-style-span style=3D"BORDER-COLLAPSE: = collapse">Hi=20 all: <DIV><BR></DIV> <DIV>Could anyone point me to the best algorithms available to estimate = SNR from=20 noisy data?</DIV> <DIV><BR></DIV> <DIV>Thanks</DIV> <DIV>Tarun</DIV> <DIV>Senior Research Engineer</DIV> <DIV>Think A Move, Ltd.</DIV></SPAN></BODY></HTML> ------_=_NextPart_001_01C9CDE2.221F986E--