Organisation internationale de normalisation

FFT Harmonic Transposer

The Chair noted he believes that there is consensus to adopt the WMF harmonic transposer, and that the only open issue is whether to retain the FFT harmonic transposer.

Kristofer Kjörling, Dolby, gave a presentation to summarize his view of the relevant issues concerning retaining the FFT harmonic transposer. Concerning complexity, his points were

• 
  For a low-rate service the choice of transposer affects the required platform peak computing power substantially (approximately 50% increase in complexity).
  
• 
  For a high-rate service the choice of transposer does not affect the required platform peak computing power.
  

Concerning quality, he used spectrograms to point out the reasons, in terms of coding noise, that the FFT transposer was graded better than the QMF transposer.

Bernhard Grill, FhG, advised the group to take the lesson of AAC backward adaptive predictor. This made an important difference in a critical signal, but ultimately was found only in AAC Main Profile.

Werner Oomen, Philips, commented that the claimed additional benefit of the FFT transposer over the QMF transposer in terms of quality is very marginal and its high complexity does not warrant keeping the FFT transposer. He advised the group to be bold and remove tools that do not demonstrate sufficient performance. Since Philips was the only company voting negative, after some discussion, Werner Oomen agreed to change his position for the sake of unanimity.

It was the consensus of the Audio subgroup to incorporate the QMF transposer technology that is proposed in contribution m19301, “Finalization of CE on QMF based harmonic transposer” into the USAC DIS and to retain the FFT transposer technology in the USAC DIS.