Hi, *"That's in general not very reliable. You can get PEAQ to say all sorts of silly things." Can you provide me links for any more effective tools other than PEAQ? Which is more reliable for Speex resampler? * *"strongly suspect that it's just not compensating for the delay introduced by the resampler. Because higher quality means higher delay, you'd find that PEAQ doesn't like it when you increase the resampler quality. You can easily remove the delay (well, not remove it but skip the zeros it produces) by calling speex_resampler_skip_zeros() once, before you start the processing." * *When I use speex_resampler_skip_zeros() for 44100 to 48000 conversion, I got a very bad quality value when I used PEAQ. Do you suggest me not to use this function for fractional rate conversions? Regards, * On Feb 15, 2008 2:23 AM, Jean-Marc Valin <jean-marc.valin@usherbrooke.ca> wrote:> Premkiran Mannava a ?crit : > > I just built a sample application with speex resampler in linux and I > > tried to resample 8K sine wave tone mono to 48k using > > speex_resample_process_int. I am using a tool called EAQUAL for audio > > quality. > > That's in general not very reliable. You can get PEAQ to say all sorts > of silly things. > > > I find the quality of Speex resampler to be decreasing when I > > increase the quality q of the resampler init function. Can some one > > give me pointers regarding this?? > > I strongly suspect that it's just not compensating for the delay > introduced by the resampler. Because higher quality means higher delay, > you'd find that PEAQ doesn't like it when you increase the resampler > quality. You can easily remove the delay (well, not remove it but skip > the zeros it produces) by calling speex_resampler_skip_zeros() once, > before you start the processing. > > > As per the API, if the quality factor > > is increased, it is assumed that the quality should increase which is no > > the way I observed with my console application > > Did you actually *listen* to the result? > > > Can some one give me pointers to any doc available on the algorithm of > > Speex resampler? > > The resampler.c file (get a recent version) has all the info/links I > have on that algo. > > Jean-Marc >-- " Excellence is not an act....but a habit..." -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20080218/5ef0a555/attachment-0001.html
On 2/18/08, Premkiran Mannava <loverays@gmail.com> wrote:> > Hi, > > *"That's in general not very reliable. You can get PEAQ to say all sorts > of silly things." > > Can you provide me links for any more effective tools other than PEAQ? > Which is more reliable for Speex resampler?*I can already tell you what Jean-Marc will say -- use your ears :) Manual testing is the best way to go for doing actual quality determining. To prevent regressions though, a tool like PEAQ might be useful -- not so much the actual quality level it determines, but the value changing a lot could then get you to do another manual test. *"strongly suspect that it's just not compensating for the delay> introduced by the resampler. Because higher quality means higher delay, > you'd find that PEAQ doesn't like it when you increase the resampler > quality. You can easily remove the delay (well, not remove it but skip > the zeros it produces) by calling speex_resampler_skip_zeros() once, > before you start the processing." > * > *When I use speex_resampler_skip_zeros() for 44100 to 48000 conversion, I > got a very bad quality value when I used PEAQ. Do you suggest me not to use > this function for fractional rate conversions?*Doesn't this tell you that PEAQ doesn't work well with speex? -- Keith Kyzivat SIPez LLC. SIP VoIP, IM and Presence Consulting http://www.SIPez.com tel: +1 (617) 273-4000 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20080218/9bb99de2/attachment.html
Thanks for your inputs. I quickly tested manually using my golden ears and I find the quality of perception is perfect :) Though I get a very bad PEAQ test value a manual test seems to be better. But I am also looking for any kind of objective measurement to validate my test appl. Is it not possible to objectively validate Speex resampler using any kind of code or tool (may not be a PEAQ based tool)? I have few more questions.... 1. Is it required for speex resampler input buffer length to be multiple of SpeexResamplerState's frac_advance value? I find if above is not the case, samples are missing in fractional rate conversion. 2. Out of resampler_basic_direct_single and resampler_basic_interpolation_single, which of them is nore reliable one? Can I use resampler_basic_direct_single for all purposes as this seems to be simpler implementation. Which one would you recommend for? On Feb 18, 2008 8:54 PM, Keith Kyzivat <kkyzivat@sipez.com> wrote:> > > On 2/18/08, Premkiran Mannava <loverays@gmail.com> wrote: > > > > Hi, > > > > *"That's in general not very reliable. You can get PEAQ to say all sorts > > of silly things." > > > > Can you provide me links for any more effective tools other than PEAQ? > > Which is more reliable for Speex resampler?* > > > I can already tell you what Jean-Marc will say -- use your ears :) > Manual testing is the best way to go for doing actual quality determining. > To prevent regressions though, a tool like PEAQ might be useful -- not so > much the actual quality level it determines, but the value changing a lot > could then get you to do another manual test. > > *"strongly suspect that it's just not compensating for the delay > > introduced by the resampler. Because higher quality means higher delay, > > you'd find that PEAQ doesn't like it when you increase the resampler > > quality. You can easily remove the delay (well, not remove it but skip > > the zeros it produces) by calling speex_resampler_skip_zeros() once, > > before you start the processing." > > * > > *When I use speex_resampler_skip_zeros() for 44100 to 48000 conversion, > > I got a very bad quality value when I used PEAQ. Do you suggest me not to > > use this function for fractional rate conversions?* > > > Doesn't this tell you that PEAQ doesn't work well with speex? > > -- > Keith Kyzivat > > SIPez LLC. > SIP VoIP, IM and Presence Consulting > http://www.SIPez.com > tel: +1 (617) 273-4000-- " Excellence is not an act....but a habit..." -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20080218/931b845d/attachment.htm