Filter Coefficients in ATRAC? Options

[pat]

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?

[Gabriel]

Do you have the specifications of the Atrac format?

[pat]

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.

[Gabriel]

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

[pat]

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.....

[Gabriel]

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.

[wkwai]

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..

[menno]

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

[pat]

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????

[wkwai]

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.

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..