Hi
Any help is appreciate!
These days I studied ogg vorbis encoder with sample rate 44.1kbsp,and I thought
the whole chart was as follows:
1. read the PCM data and store them in a buffer
2. do MDCT on those PCM data
3. do FFT on those PCM data
4. use both the FFT and MDCT coefficient to the psychoacustic modle and derive
an maximum allowable noise curve,
and during this curse ,do the window anlysis and decide which window to use
for the next coming PCM data
5. use the psychoacustic result on the MDCT coefficient and remove those below
the maximum allowable noise curve
6. do floor anlysis , encode the floor and pack them in the bitstream
7. do residue = MDCT/floor ,derive residue curve and choose one type of residue
encoder method and use huffman tree
to pack them in the bitstream(use stereo coupling when multiple channels).
8. if it is not the end of the coming PCM data,do the loop again.
For the first time,we use a short window ,after that, we decide the type of
window according to the psychoacustic model.
Is it all right ?
Looking forward for your letter!
Best regards
??????? Lisa
????????fangchao_dlut@sina.com
??????????2005-09-27