Re: [FFmpeg-devel] [PATCH 1/2] avcodec/vorbisenc: Add pre-echo detection

[Tyler Jones]

On Fri, Jul 14, 2017 at 12:43:26AM +0200, Michael Niedermayer wrote:
> On Wed, Jul 12, 2017 at 04:18:06PM -0600, Tyler Jones wrote:
> [...]
> 
> 
> > +av_cold int psy_vorbis_init(VorbisPsyContext *vpctx, int sample_rate,
> > +int channels, int blocks)
> 
> non static functions needs a prefix
> in this case ff_
> 
> 
> [...]
> 
> -- 
> Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
> 
> Does the universe only have a finite lifespan? No, its going to go on
> forever, its just that you wont like living in it. -- Hiranya Peiri

Noted, thank you for noticing. This is changed for the next revision.

Tyler Jones


signature.asc
Description: PGP signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

Re: [FFmpeg-devel] [PATCH 1/2] avcodec/vorbisenc: Add pre-echo detection

[Michael Niedermayer]

On Wed, Jul 12, 2017 at 04:18:06PM -0600, Tyler Jones wrote:
[...]


> +av_cold int psy_vorbis_init(VorbisPsyContext *vpctx, int sample_rate,
> +int channels, int blocks)

non static functions needs a prefix
in this case ff_


[...]

-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Does the universe only have a finite lifespan? No, its going to go on
forever, its just that you wont like living in it. -- Hiranya Peiri


signature.asc
Description: Digital signature
___
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel

[FFmpeg-devel] [PATCH 1/2] avcodec/vorbisenc: Add pre-echo detection

[Tyler Jones]

The encoder will attempt to determine the existence of transient
signals by applying a 4th order highpass filter to remove dominant
low frequency waveforms. Frames are then split up into blocks
where the variance is calculated and compared with blocks from
the previous frame. A preecho is only likely to be noticeable when
relatively quiet audio is followed by a loud transient signal.

Signed-off-by: Tyler Jones 
---
 libavcodec/Makefile|   2 +-
 libavcodec/vorbisenc.c |  28 +++--
 libavcodec/vorbispsy.c | 153 +
 libavcodec/vorbispsy.h |  79 +
 4 files changed, 256 insertions(+), 6 deletions(-)
 create mode 100644 libavcodec/vorbispsy.c
 create mode 100644 libavcodec/vorbispsy.h

diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index b440a00..2db6727 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -611,7 +611,7 @@ OBJS-$(CONFIG_VMNC_DECODER)+= vmnc.o
 OBJS-$(CONFIG_VORBIS_DECODER)  += vorbisdec.o vorbisdsp.o vorbis.o \
   vorbis_data.o
 OBJS-$(CONFIG_VORBIS_ENCODER)  += vorbisenc.o vorbis.o \
-  vorbis_data.o
+  vorbis_data.o vorbispsy.o
 OBJS-$(CONFIG_VP3_DECODER) += vp3.o
 OBJS-$(CONFIG_VP5_DECODER) += vp5.o vp56.o vp56data.o vp56rac.o
 OBJS-$(CONFIG_VP6_DECODER) += vp6.o vp56.o vp56data.o \
diff --git a/libavcodec/vorbisenc.c b/libavcodec/vorbisenc.c
index bf21a3b..3482cf0 100644
--- a/libavcodec/vorbisenc.c
+++ b/libavcodec/vorbisenc.c
@@ -33,6 +33,7 @@
 #include "mathops.h"
 #include "vorbis.h"
 #include "vorbis_enc_data.h"
+#include "vorbispsy.h"
 
 #include "audio_frame_queue.h"
 #include "libavfilter/bufferqueue.h"
@@ -136,6 +137,7 @@ typedef struct vorbis_enc_context {
 int64_t next_pts;
 
 AVFloatDSPContext *fdsp;
+VorbisPsyContext *vpctx;
 } vorbis_enc_context;
 
 #define MAX_CHANNELS 2
@@ -272,11 +274,12 @@ static int create_vorbis_context(vorbis_enc_context *venc,
 vorbis_enc_floor   *fc;
 vorbis_enc_residue *rc;
 vorbis_enc_mapping *mc;
-int i, book, ret;
+int i, book, ret, blocks;
 
 venc->channels= avctx->channels;
 venc->sample_rate = avctx->sample_rate;
-venc->log2_blocksize[0] = venc->log2_blocksize[1] = 11;
+venc->log2_blocksize[0] = 8;
+venc->log2_blocksize[1] = 11;
 
 venc->ncodebooks = FF_ARRAY_ELEMS(cvectors);
 venc->codebooks  = av_malloc(sizeof(vorbis_enc_codebook) * 
venc->ncodebooks);
@@ -464,6 +467,12 @@ static int create_vorbis_context(vorbis_enc_context *venc,
 if ((ret = dsp_init(avctx, venc)) < 0)
 return ret;
 
+blocks = 1 << (venc->log2_blocksize[1] - venc->log2_blocksize[0]);
+venc->vpctx = av_mallocz(sizeof(VorbisPsyContext));
+if (!venc->vpctx || (ret = psy_vorbis_init(venc->vpctx, venc->sample_rate,
+   venc->channels, blocks)) < 0)
+return AVERROR(ENOMEM);
+
 return 0;
 }
 
@@ -1071,22 +1080,23 @@ static void move_audio(vorbis_enc_context *venc, int 
sf_size)
 float *save = venc->saved + ch * frame_size;
 const float *input = (float *) cur->extended_data[ch];
 const size_t len  = cur->nb_samples * sizeof(float);
-
 memcpy(offset + sf*sf_size, input, len);
 memcpy(save + sf*sf_size, input, len);   // Move samples for next 
frame
 }
 av_frame_free();
 }
 venc->have_saved = 1;
-memcpy(venc->scratch, venc->samples, 2 * venc->channels * frame_size);
+memcpy(venc->scratch, venc->samples, sizeof(float) * venc->channels * 2 * 
frame_size);
 }
 
 static int vorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
const AVFrame *frame, int *got_packet_ptr)
 {
 vorbis_enc_context *venc = avctx->priv_data;
-int i, ret, need_more;
+int i, ret, need_more, ch;
+int curr_win = 1;
 int frame_size = 1 << (venc->log2_blocksize[1] - 1);
+int block_size = 1 << (venc->log2_blocksize[0] - 1);
 vorbis_enc_mode *mode;
 vorbis_enc_mapping *mapping;
 PutBitContext pb;
@@ -1121,6 +1131,13 @@ static int vorbis_encode_frame(AVCodecContext *avctx, 
AVPacket *avpkt,
 
 move_audio(venc, avctx->frame_size);
 
+for (ch = 0; ch < venc->channels; ch++) {
+float *scratch = venc->scratch + 2 * ch * frame_size + frame_size;
+
+if (!psy_vorbis_block_frame(venc->vpctx, scratch, ch, frame_size, 
block_size))
+curr_win = 0;
+}
+
 if (!apply_window_and_mdct(venc))
 return 0;
 
@@ -1252,6 +1269,7 @@ static av_cold int vorbis_encode_close(AVCodecContext 
*avctx)
 ff_mdct_end(>mdct[1]);
 ff_af_queue_close(>afq);
 ff_bufqueue_discard_all(>bufqueue);
+psy_vorbis_close(venc->vpctx);
 
 av_freep(>extradata);
 
diff --git