VorbisHi Any help is appreciate. I have two questions. First, I looked into the vorbis-?-specific and was confused by the floor1 algorithm. I think at last the aim is to derive the piecewise curve with the list X and Y,then when encodering ,why not use the selected point orderly to get the curve ? In other words, the specific use the list of [0,128,64,32,96,16,48,80] , why can not use [0,16,32,48,64,80,96,112,128] and encoder the diffence between the Y and the previous Y directly? Second, from the previos letter,I know that floor is the least squared error approximation of the maximum amount of noise allowable for each mdct coefficient/frequency (determined by the psychoacustic model). Now I am just confused whatever the floor really is .Is it a predition to the MDCT coefficient or just an estimation to the maximum allowable noise? And how the psychoaustic result is applied to the compution of the floor? Looking forward to your letter! ????????Lisa ????????fangchao_dlut@sina.com ??????????2005-09-26
Hi! h?tf? 26 szeptember 2005 04.51-kor fangchao ezeket a bolcs gondolatokat fogalmazta meg:> Second, from the previos letter,I know that floor is the least squared > error approximation of the maximum amount of noise allowable for each mdct > coefficient/frequency (determined by the psychoacustic model). Now I am > just confused whatever the floor really is .Is it a predition to the MDCT > coefficient or just an estimation to the maximum allowable noise?The email "Re: [Vorbis] Floor 1 Decode" from 16 september on this list might help.> And how the psychoaustic result is applied to the compution of the floor?IMHO the psychoacustic result is the allowable noise for each mdct coefficient/frequency.> Looking forward to your letter! > > > ????????Lisa > ????????fangchao_dlut@sina.com > ??????????2005-09-26bye Denes -- --- What kills me, doesn't make me stronger.
On Mon, Sep 26, 2005 at 01:36:47PM +0200, Balatoni Denes wrote:> Hi! > > h?tf?? 26 szeptember 2005 04.51-kor fangchao ezeket a bolcs gondolatokat > fogalmazta meg: > > Second, from the previos letter,I know that floor is the least squared > > error approximation of the maximum amount of noise allowable for each mdct > > coefficient/frequency (determined by the psychoacustic model). Now I am > > just confused whatever the floor really is .Is it a predition to the MDCT > > coefficient or just an estimation to the maximum allowable noise? > > The email "Re: [Vorbis] Floor 1 Decode" from 16 september on this list might > help. > > > And how the psychoaustic result is applied to the compution of the floor? > > IMHO the psychoacustic result is the allowable noise for each mdct > coefficient/frequency.No; vorbis does not track allowable noise (this strategy was used by original mp3, but was a big red herring). The floor is intended to represent desired unit-scale in a band, in essence, the desired amount of resolution. During encode, the spectrum is quantized according to the floor and then normalized in an effort to preserve tone and noise energy over any given critical band. Monty
> First, I looked into the vorbis-$B*!(B-specific and was confused by the > floor1 algorithm. > I think at last the aim is to derive the piecewise curve with the list > X and Y,then when encodering ,why not > use the selected point orderly to get the curve ? In other words, > the specific use the list of [0,128,64,32,96,16,48,80] , > why can not use [0,16,32,48,64,80,96,112,128] and encoder the diffence > between the Y and the previous Y directly?Because that way, more bits would be needed to encode the same data. Segher