Colin McQuillan wrote:
2009/6/2 Robert Swain <[email protected]>:
cmcq wrote:
Author: cmcq
Date: Tue Jun 2 13:58:23 2009
New Revision: 4376
Log:
Rescale AMR samples from 16-bit PCM values
The samples /A-codecs/amr/{sample,sample2,whatireallywant}.amr now produce
recognizable output. There is still significant distortion.
Great! :D
Modified:
amr/amrnbfloatdata.h
amr/amrnbfloatdec.c
Modified: amr/amrnbfloatdata.h
==============================================================================
--- amr/amrnbfloatdata.h Mon Jun 1 19:34:02 2009 (r4375)
+++ amr/amrnbfloatdata.h Tue Jun 2 13:58:23 2009 (r4376)
@@ -38,6 +38,9 @@
#define AMR_BLOCK_SIZE 160
#define AMR_SUBFRAME_SIZE 40
+// AMR is designed to produce 16-bit PCM samples (3GPP TS 26.090 4.2)
+#define AMR_SAMPLE_SCALE 32768.0
Maybe have:
#define AMR_SAMPLE_BOUND 32768.0
#define AMR_SAMPLE_SCALE 1.0/32768.0
and edit the other code accordingly so that you can use * rather than /. We
may have some suitable dsputil function for this anyway so it's not
massively important. Michael would have a suggestion about it later I
expect.
Since most of the processes are linear, I should be able to scale the
innovative vectors definitions in amrnbfloatdata.h. However, I need to
investigate what other effects this will have.
OK.
Do you have any scripts to help generate amrnbfloatdata.h except reformat.pl?
I'll have a look around but I think that's all I used and just edited it
accordingly.
Regards,
Rob
_______________________________________________
FFmpeg-soc mailing list
[email protected]
https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-soc
