topic Re: Bug in AMR-WB codec in Intel® Integrated Performance Primitives

Bug in AMR-WB codec

vadim1 — Tue, 13 Dec 2005 02:24:35 GMT

I think, i've stumbled on buf in AMR-WB codec in IPP V5.0

In the funciongetBitstreamSize in the file uscamrwb.c

The function calculates the wron number of bytes per AMR frame:
it contains following line:

nbytes = ((BitsLenTbl[usedRate] + 14) / 15) * sizeof(short);

I believe it should be:

a)
nbytes = ((BitsLenTbl[usedRate] + 15) / 16) * sizeof(short);

Or maybe:
b)
nbytes = ((BitsLenTbl[usedRate] + 7) / 8) * sizeof(short);

the b) is equivqlent to what is done in gsmamr (amr narrowband) module.

Re: Bug in AMR-WB codec

Vyacheslav_Baranniko — Tue, 13 Dec 2005 19:33:29 GMT

Thanks a lot!

This is a bug. The bitstream length shouldbe calculated either wayyou pointed out. The bugwill be fixed in thenearest future version of IPP.

Kind regards

Vyacheslav

Re: Bug in AMR-WB codec

vadim1 — Tue, 13 Dec 2005 20:55:08 GMT

Actually thinkink some more about it
c) is buggy too

the correct code will be:

d)
nbytes = ((BitsLenTbl[usedRate] + 7) / 8);

and b) is faulty too because it will always generate an even
result which is incorrect

Vadim