overall level of the masking threshold (#ARIJIT BISWAS# )

Subject: overall level of the masking threshold From: #ARIJIT BISWAS# <arijit17@xxxxxxxx> Date: Tue, 21 Feb 2006 21:46:05 +0800 This is a multi-part message in MIME format. ------_=_NextPart_001_01C636ED.296C89FB Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear Daniel & Alexander: =20 Thanks a lot for nice opinions, and apologies for my slow reaction. It = was really nice and helpful.=20 =20 In the meantime, I have more questions for the list: =20 I was wondering if it is possible to predict the "overall level of the = masking threshold" based on some statistics of the input signal (e.g. = power, power spectral density, etc). By "overall level of the masking = threshold", I mean the average of the masking threshold over frequency.=20 For example, the threshold in quiet has an overall level of ~ 32 dB =20 It would be great to have some kind of relation between the overall = level of the masking threshold with the level of the input signal and = the level of the threshold in quiet. Is there any formulas existing somewhere in the psychoacoustics = literatures? =20 =20 Any suggestions and/or comments in this regard will be highly = appreciated. =20 Thanks and regards, ~Arijit=20 =20 =20 ________________________________ From: AUDITORY Research in Auditory Perception on behalf of Danijel = Domazet Sent: Thu 2/9/2006 11:55 AM To: AUDITORY@xxxxxxxx Subject: Re: computational complexity of psychoacoustic models Hi Arijit, Try avoiding MPEG psychoacoustic model 2 - I think it is too complex. There are a few things that are important when desingning SIMPLE psycho model: - try to avoid separate time/freq. transformation (usualy FFT). Use the result of the one that is already present in the encoder (MDCT most = likely). It isn't as good but spares you a FFT computation. Results are more than acceptable. - don't define separeate critical bands like in Psycho 2 (that better = fit human hearing), use the ones defined in your encoder as scalefactor = bands, it will be much simpler. - tonality estimation might also be unnecessary. Just assume the = constant masking for tonal and non-tonal singnals, it will do the job for most signals (you might loose some quality for strong tonal samples but it = might not be too critical). - if you have to include tonality detection - don't calculate it based = on prediction accross frames, lookahead buffers will increase the delay and complexity also. MPEG psycho model 2 has some really unnecessary = lookaheads. Use some other method for tonality estimation (Spectral Flatness Measure = for example). - don't complicate with the spreading function, simple triangular = function will do the job. - detect transients in TIME domain. - estimate scalefactors directly from masking threasholds, don't use inner-and-outter loop method like Psycho 2 recommends (many iterations = slow you down drastically). What I would do is somehting like: - calculate time/freq transformation - calculate energy accros sritical bands - calculate masking (or use constant) - calculate masking threshold as energy * masking - apply spreading function - apply threashold in quiet (this will give you the main result of the psycho analysis - the masking threashold) - convert masking thresholds directly to scalefactors If your quantized spectar doesn't fit the bitrate, just increment ALL scalefacotors at the same time and repeat the quantization. I hope this helped. It you don't understand all this now, don't worry - = you will when you get involved with psychoacoustics some more. Also, take a look at the psychoacoustic model of the Enhanced aacPlus general audio codec from 3GPP - TS 26.403. Regards, Daniel ----- Original Message ----- From: "alexander lerch" <lerch@xxxxxxxx> To: <AUDITORY@xxxxxxxx> Sent: Wednesday, February 08, 2006 1:49 PM Subject: Re: [AUDITORY] computational complexity of psychoacoustic = models The choice is, at least for all MPEG codecs, completely up to the developer. You can decide not to use a psychoacoustic model at all, or you can decide to use a complex model to gain as much quality as = possible. Oftenly used steps are: FFT Critical Band grouping Conversion to dB (Analysis of tonality of possible maskers) calculation of masking threshold via masking model Have a look at the psychoacoustic model 2 in the informative part of the MPEG-1 standard. Kind regards, Alexander #ARIJIT BISWAS# wrote: > Hi List: > > > > I'm interested to know the computational complexity (number of = additions > and multiplications) of psychoacoustic models used in audio coding. > > Well, to be more specific, let's say if I'm targeting to build a = "fast" > psychoacoustic model, which existing model and/or what kind of > computational complexity should I try to beat? > > > > Any help/suggestions/references in this direction will be highly > appreciated. > > > > Best Regards, > > ~Arijit > -- dipl. ing. alexander lerch zplane.development :www.zplane.de katzbachstr.21 d-10965 berlin fon: +49.30.854 09 15.0 fax: +49.30.854 09 15.5 ------_=_NextPart_001_01C636ED.296C89FB Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable <META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; = charset=3Diso-8859-1">=0A= <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">=0A= <HTML>=0A= <HEAD>=0A= =0A= <META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version = 6.5.7638.1">=0A= <TITLE>Re: computational complexity of psychoacoustic models</TITLE>=0A= </HEAD>=0A= <BODY>=0A= <DIV id=3DidOWAReplyText35681 dir=3Dltr>=0A= Dear Daniel & Alexander:<?xml:namespace prefix =3D o = ns =3D =0A= "urn:schemas-microsoft-com:office:office" /><o:p></o:p>=0A=  <o:p></o:p>=0A= Thanks a lot for nice opinions, and apologies for my = slow =0A= reaction. It was really nice and helpful. <o:p></o:p>=0A=  <o:p></o:p>=0A= In the meantime, I have more questions for the =0A= list:<o:p></o:p>=0A=  <o:p></o:p>=0A= I was wondering if it is possible to predict the = “overall level of =0A= the masking threshold” based on some statistics of the input = signal (e.g. power, =0A= power spectral density, etc). By “overall level of the masking = threshold”, I =0A= mean the average of the masking threshold over frequency. =0A= <o:p></o:p>=0A= For example, the threshold in quiet has an overall level = of  ~ 32 = dB<o:p></o:p>=0A=  <o:p></o:p>=0A= It would be great to have some kind of relation between = the =0A= overall level of the masking threshold with the level of the input = signal and =0A= the level of the threshold in quiet.=0A= <o:p>Is there any formulas existing somewhere in the = psychoacoustics =0A= literatures?  <o:p></o:p></o:p>=0A=  <o:p></o:p>=0A= Any suggestions and/or comments in this regard will be = highly =0A= appreciated.<o:p></o:p>=0A=  <o:p></o:p>=0A= Thanks and regards,<o:p></o:p>=0A= <DIV dir=3Dltr><SPAN lang=3DEN-US =0A= style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Verdana; mso-ansi-language: = EN-US; mso-fareast-font-family: 'Times New Roman'; mso-fareast-language: = EN-US; mso-bidi-language: AR-SA; mso-bidi-font-size: 12.0pt; = mso-bidi-font-family: 'Times New Roman'">~Arijit </DIV>=0A= <DIV dir=3Dltr><SPAN lang=3DEN-US =0A= style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Verdana; mso-ansi-language: = EN-US; mso-fareast-font-family: 'Times New Roman'; mso-fareast-language: = EN-US; mso-bidi-language: AR-SA; mso-bidi-font-size: 12.0pt; = mso-bidi-font-family: 'Times New Roman'"> </DIV>=0A= <DIV dir=3Dltr><SPAN lang=3DEN-US =0A= style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Verdana; mso-ansi-language: = EN-US; mso-fareast-font-family: 'Times New Roman'; mso-fareast-language: = EN-US; mso-bidi-language: AR-SA; mso-bidi-font-size: 12.0pt; = mso-bidi-font-family: 'Times New Roman'"> </DIV></DIV></SP= AN>=0A= <DIV dir=3Dltr> =0A= <HR tabIndex=3D-1>=0A= From: AUDITORY Research in Auditory = Perception =0A= on behalf of Danijel Domazet Sent: Thu 2/9/2006 11:55 = AM To: =0A= AUDITORY@xxxxxxxx Subject: Re: computational complexity = of =0A= psychoacoustic models </DIV>=0A= <DIV>=0A= Hi Arijit, Try avoiding MPEG psychoacoustic model 2 = - I think =0A= it is too complex. There are a few things that are important = when =0A= desingning SIMPLE psycho model: - try to avoid separate = time/freq. =0A= transformation (usualy FFT). Use the result of the one that is = already =0A= present in the encoder (MDCT most likely). It isn't as good but = spares you a =0A= FFT computation. Results are more than acceptable. - don't = define =0A= separeate critical bands like in Psycho 2 (that better fit human = hearing), =0A= use the ones defined in your encoder as scalefactor bands, it will be = much =0A= simpler. - tonality estimation might also be unnecessary. Just = assume the =0A= constant masking for tonal and non-tonal singnals, it will do the job = for =0A= most signals (you might loose some quality for strong tonal samples = but it =0A= might not be too critical). - if you have to include tonality =0A= detection - don't calculate it based on prediction accross frames, = lookahead =0A= buffers will increase the delay and complexity also. MPEG psycho = model 2 has =0A= some really unnecessary lookaheads. Use some other method for = tonality =0A= estimation (Spectral Flatness Measure for example). - don't = complicate =0A= with the spreading function, simple triangular function will do the =0A= job. - detect transients in TIME domain. - estimate = scalefactors =0A= directly from masking threasholds, don't use inner-and-outter loop = method =0A= like Psycho 2 recommends (many iterations slow you down =0A= drastically). What I would do is somehting like: - = calculate =0A= time/freq transformation - calculate energy accros sritical = bands - =0A= calculate masking (or use constant) - calculate masking threshold as = energy * =0A= masking - apply spreading function - apply threashold in quiet = (this will =0A= give you the main result of the psycho analysis - the masking =0A= threashold) - convert masking thresholds directly to = scalefactors If your =0A= quantized spectar doesn't fit the bitrate, just increment = ALL scalefacotors =0A= at the same time and repeat the quantization. I hope this = helped. It =0A= you don't understand all this now, don't worry - you will when you = get =0A= involved with psychoacoustics some more. Also, take a look at the =0A= psychoacoustic model of the Enhanced aacPlus general audio codec from = 3GPP - =0A= TS 26.403. Regards, Daniel ----- Original = Message =0A= ----- From: "alexander lerch" <lerch@xxxxxxxx> To: =0A= <AUDITORY@xxxxxxxx> Sent: Wednesday, February 08, 2006 = 1:49 =0A= PM Subject: Re: [AUDITORY] computational complexity of psychoacoustic =0A= models The choice is, at least for all MPEG codecs, = completely up to =0A= the developer. You can decide not to use a psychoacoustic model at = all, =0A= or you can decide to use a complex model to gain as much quality as =0A= possible. Oftenly used steps are: FFT Critical Band =0A= grouping Conversion to dB (Analysis of tonality of possible =0A= maskers) calculation of masking threshold via masking = model Have a =0A= look at the psychoacoustic model 2 in the informative part of = the MPEG-1 =0A= standard. Kind regards, Alexander #ARIJIT BISWAS# =0A= wrote: > Hi List: > > > > I’m = interested to know =0A= the computational complexity (number of additions > and = multiplications) =0A= of psychoacoustic models used in audio coding. > > Well, to = be more =0A= specific, let’s say if I’m targeting to build a = “fast” > psychoacoustic =0A= model, which existing model and/or what kind of > computational = complexity =0A= should I try to beat? > > > > Any =0A= help/suggestions/references in this direction will be highly > =0A= appreciated. > > > > Best = Regards, > > =0A= ~Arijit > -- dipl. ing. alexander =0A= lerch zplane.development :www.zplane.de katzbachstr.21 d-= 10965 =0A= berlin fon: +49.30.854 09 15.0 fax: +49.30.854 09 =0A= 15.5 </DIV>=0A= =0A= </BODY>=0A= </HTML> ------_=_NextPart_001_01C636ED.296C89FB--

This message came from the mail archive
http://www.auditory.org/postings/2006/
maintained by:

DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University