Since there is a standard filter coefficients given in mp3(ISO 11172-3), I wonder if there is a standard filter coefficients I can find for ATRAC?

If there is no, how do i go about deciding the coefficients i'm going to use to build the ATRAC encoder?

Do you have the specifications of the Atrac format?

No. I don't have something as detailed like the ISO 11172. So far I only have the overall block diagrams, without the detailed coefficients.

I only know the QMF have to divide the input signal into 3 subbands, 0 - 5.5kHz, 5.5 - 11kHz, and 11 - 22kHz.

So how could you be able to write an encoder without specifications or reference source code?

Hm.....that's my problem. I'm actually doing this as my project for part of my coursework.....

Is there anyone who is familiar with ATRAC? I sincerely need help.....

The problem is that we do not have access to Atrac specifications (perhaps you have them), so most of us are unable to help you on this matter.

Sony guards its commercial secrets very well.... No doubt that without the proper specs, you can't implement an ATRAC coder/decoder.. Even the huffman tables are unique.. even if it does not mention any specific quantization algorithm..

Filters are not really the problem. If you know the subband edges you can easily design them yourself
Try searching for a book/paper on subband filter design.

BTW: I think ATRAC does 2 filterbanks. One that splits the signal in 2 subbands (0-11,11-22) and one that splits the first subband in 2 more (0-5.5,5.5-11). You can do the second filterbank by decimating the prototype filter of the first by 2. This leaves you to design just a prototype filter for the first filterbank that has only 2 subbands (so splitting signal in halve). This is the easiest filterbank possible

Menno

Thanks guys. Think I still manage to handle the filter. But then when come to the Bit Allocation and the Spectral Quantization parts, i become more n more headache. As what Gabriel said, Sony guides their technology tightly that I couldn't find any of the standards carry on with my work. So i am kinda stuck now.

Does anyone can help me out????

I think it is not as simple as that.. The reconstruction subband filterbanks MUST match the analysis subband filterbanks, otherwise, there could be aliasing artifacts in the decoded audio data..