Hi All,

Comparing the tonality calculation model of ISO MPEG-4(Psy 2), it appears that Johnston's tonality calculation(SFM) is only local to the current frame and Ivan's point (Improved ISO AAC Encoder) on having some sort of intraframe tonality estimation model is rational.

I have implemented the model using two methods and would like your opinion on that,
In first method I used,
1 - Gm *= ((Energy[b] + moving average_of_Energy[b])/2), and similarly Am
and in second method I used,
2 - Mean of Previous Gm and current Gm, and similarly Am

When I look at the tonality calculation using unpredictabilty in ISO-14496-3, I feel my technique is much simpler and feel the need of improvement.

I would like to know of other methods to calculate the Intraframe SFM or if you know of any published resource for me to look at.

Regards,
Nabsha

Well, since Johnston wrote both Psychoacoustic Model 2 and the paper using SFM, and he wrote the Psychoacoustic Model 2 later, I would think the answer would be obvious.