[FFmpeg-cvslog] g723_1: add comfort noise generation
Kostya Shishkov
git at videolan.org
Fri Aug 31 13:08:35 CEST 2012
ffmpeg | branch: master | Kostya Shishkov <kostya.shishkov at gmail.com> | Thu Aug 23 19:51:11 2012 +0200| [04fc5c6bde5da842e8b40a9a76eb72c8327ae40b] | committer: Kostya Shishkov
g723_1: add comfort noise generation
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=04fc5c6bde5da842e8b40a9a76eb72c8327ae40b
---
libavcodec/g723_1.c | 226 +++++++++++++++++++++++++++++++++++++++++--
libavcodec/g723_1_data.h | 6 ++
tests/fate/voice.mak | 6 ++
tests/ref/fate/g723_1-dec-7 | 12 +++
tests/ref/fate/g723_1-dec-8 | 121 +++++++++++++++++++++++
5 files changed, 364 insertions(+), 7 deletions(-)
diff --git a/libavcodec/g723_1.c b/libavcodec/g723_1.c
index 793b2d3..03dd62d 100644
--- a/libavcodec/g723_1.c
+++ b/libavcodec/g723_1.c
@@ -35,6 +35,8 @@
#include "celp_filters.h"
#include "g723_1_data.h"
+#define CNG_RANDOM_SEED 12345
+
/**
* G723.1 frame types
*/
@@ -84,6 +86,7 @@ typedef struct g723_1_context {
int erased_frames;
int16_t prev_lsp[LPC_ORDER];
+ int16_t sid_lsp[LPC_ORDER];
int16_t prev_excitation[PITCH_MAX];
int16_t excitation[PITCH_MAX + FRAME_LEN + 4];
int16_t synth_mem[LPC_ORDER];
@@ -91,6 +94,7 @@ typedef struct g723_1_context {
int iir_mem[LPC_ORDER];
int random_seed;
+ int cng_random_seed;
int interp_index;
int interp_gain;
int sid_gain;
@@ -99,7 +103,7 @@ typedef struct g723_1_context {
int pf_gain;
int postfilter;
- int16_t audio[FRAME_LEN + LPC_ORDER + PITCH_MAX];
+ int16_t audio[FRAME_LEN + LPC_ORDER + PITCH_MAX + 4];
} G723_1_Context;
static av_cold int g723_1_decode_init(AVCodecContext *avctx)
@@ -116,6 +120,10 @@ static av_cold int g723_1_decode_init(AVCodecContext *avctx)
avctx->coded_frame = &p->frame;
memcpy(p->prev_lsp, dc_lsp, LPC_ORDER * sizeof(*p->prev_lsp));
+ memcpy(p->sid_lsp, dc_lsp, LPC_ORDER * sizeof(*p->sid_lsp));
+
+ p->cng_random_seed = CNG_RANDOM_SEED;
+ p->past_frame_type = SID_FRAME;
return 0;
}
@@ -983,6 +991,201 @@ static void formant_postfilter(G723_1_Context *p, int16_t *lpc,
}
}
+static int sid_gain_to_lsp_index(int gain)
+{
+ if (gain < 0x10)
+ return gain << 6;
+ else if (gain < 0x20)
+ return gain - 8 << 7;
+ else
+ return gain - 20 << 8;
+}
+
+static inline int cng_rand(int *state, int base)
+{
+ *state = (*state * 521 + 259) & 0xFFFF;
+ return (*state & 0x7FFF) * base >> 15;
+}
+
+static int estimate_sid_gain(G723_1_Context *p)
+{
+ int i, shift, seg, seg2, t, val, val_add, x, y;
+
+ shift = 16 - p->cur_gain * 2;
+ if (shift > 0)
+ t = p->sid_gain << shift;
+ else
+ t = p->sid_gain >> -shift;
+ x = t * cng_filt[0] >> 16;
+
+ if (x >= cng_bseg[2])
+ return 0x3F;
+
+ if (x >= cng_bseg[1]) {
+ shift = 4;
+ seg = 3;
+ } else {
+ shift = 3;
+ seg = (x >= cng_bseg[0]);
+ }
+ seg2 = FFMIN(seg, 3);
+
+ val = 1 << shift;
+ val_add = val >> 1;
+ for (i = 0; i < shift; i++) {
+ t = seg * 32 + (val << seg2);
+ t *= t;
+ if (x >= t)
+ val += val_add;
+ else
+ val -= val_add;
+ val_add >>= 1;
+ }
+
+ t = seg * 32 + (val << seg2);
+ y = t * t - x;
+ if (y <= 0) {
+ t = seg * 32 + (val + 1 << seg2);
+ t = t * t - x;
+ val = (seg2 - 1 << 4) + val;
+ if (t >= y)
+ val++;
+ } else {
+ t = seg * 32 + (val - 1 << seg2);
+ t = t * t - x;
+ val = (seg2 - 1 << 4) + val;
+ if (t >= y)
+ val--;
+ }
+
+ return val;
+}
+
+static void generate_noise(G723_1_Context *p)
+{
+ int i, j, idx, t;
+ int off[SUBFRAMES];
+ int signs[SUBFRAMES / 2 * 11], pos[SUBFRAMES / 2 * 11];
+ int tmp[SUBFRAME_LEN * 2];
+ int16_t *vector_ptr;
+ int64_t sum;
+ int b0, c, delta, x, shift;
+
+ p->pitch_lag[0] = cng_rand(&p->cng_random_seed, 21) + 123;
+ p->pitch_lag[1] = cng_rand(&p->cng_random_seed, 19) + 123;
+
+ for (i = 0; i < SUBFRAMES; i++) {
+ p->subframe[i].ad_cb_gain = cng_rand(&p->cng_random_seed, 50) + 1;
+ p->subframe[i].ad_cb_lag = cng_adaptive_cb_lag[i];
+ }
+
+ for (i = 0; i < SUBFRAMES / 2; i++) {
+ t = cng_rand(&p->cng_random_seed, 1 << 13);
+ off[i * 2] = t & 1;
+ off[i * 2 + 1] = ((t >> 1) & 1) + SUBFRAME_LEN;
+ t >>= 2;
+ for (j = 0; j < 11; j++) {
+ signs[i * 11 + j] = (t & 1) * 2 - 1 << 14;
+ t >>= 1;
+ }
+ }
+
+ idx = 0;
+ for (i = 0; i < SUBFRAMES; i++) {
+ for (j = 0; j < SUBFRAME_LEN / 2; j++)
+ tmp[j] = j;
+ t = SUBFRAME_LEN / 2;
+ for (j = 0; j < pulses[i]; j++, idx++) {
+ int idx2 = cng_rand(&p->cng_random_seed, t);
+
+ pos[idx] = tmp[idx2] * 2 + off[i];
+ tmp[idx2] = tmp[--t];
+ }
+ }
+
+ vector_ptr = p->audio + LPC_ORDER;
+ memcpy(vector_ptr, p->prev_excitation,
+ PITCH_MAX * sizeof(*p->excitation));
+ for (i = 0; i < SUBFRAMES; i += 2) {
+ gen_acb_excitation(vector_ptr, vector_ptr,
+ p->pitch_lag[i >> 1], &p->subframe[i],
+ p->cur_rate);
+ gen_acb_excitation(vector_ptr + SUBFRAME_LEN,
+ vector_ptr + SUBFRAME_LEN,
+ p->pitch_lag[i >> 1], &p->subframe[i + 1],
+ p->cur_rate);
+
+ t = 0;
+ for (j = 0; j < SUBFRAME_LEN * 2; j++)
+ t |= FFABS(vector_ptr[j]);
+ t = FFMIN(t, 0x7FFF);
+ if (!t) {
+ shift = 0;
+ } else {
+ shift = -10 + av_log2(t);
+ if (shift < -2)
+ shift = -2;
+ }
+ sum = 0;
+ if (shift < 0) {
+ for (j = 0; j < SUBFRAME_LEN * 2; j++) {
+ t = vector_ptr[j] << -shift;
+ sum += t * t;
+ tmp[j] = t;
+ }
+ } else {
+ for (j = 0; j < SUBFRAME_LEN * 2; j++) {
+ t = vector_ptr[j] >> shift;
+ sum += t * t;
+ tmp[j] = t;
+ }
+ }
+
+ b0 = 0;
+ for (j = 0; j < 11; j++)
+ b0 += tmp[pos[(i / 2) * 11 + j]] * signs[(i / 2) * 11 + j];
+ b0 = b0 * 2 * 2979LL + (1 << 29) >> 30; // approximated division by 11
+
+ c = p->cur_gain * (p->cur_gain * SUBFRAME_LEN >> 5);
+ if (shift * 2 + 3 >= 0)
+ c >>= shift * 2 + 3;
+ else
+ c <<= -(shift * 2 + 3);
+ c = (av_clipl_int32(sum << 1) - c) * 2979LL >> 15;
+
+ delta = b0 * b0 * 2 - c;
+ if (delta <= 0) {
+ x = -b0;
+ } else {
+ delta = square_root(delta);
+ x = delta - b0;
+ t = delta + b0;
+ if (FFABS(t) < FFABS(x))
+ x = -t;
+ }
+ shift++;
+ if (shift < 0)
+ x >>= -shift;
+ else
+ x <<= shift;
+ x = av_clip(x, -10000, 10000);
+
+ for (j = 0; j < 11; j++) {
+ idx = (i / 2) * 11 + j;
+ vector_ptr[pos[idx]] = av_clip_int16(vector_ptr[pos[idx]] +
+ (x * signs[idx] >> 15));
+ }
+
+ /* copy decoded data to serve as a history for the next decoded subframes */
+ memcpy(vector_ptr + PITCH_MAX, vector_ptr,
+ sizeof(*vector_ptr) * SUBFRAME_LEN * 2);
+ vector_ptr += SUBFRAME_LEN * 2;
+ }
+ /* Save the excitation for the next frame */
+ memcpy(p->prev_excitation, p->audio + LPC_ORDER + FRAME_LEN,
+ PITCH_MAX * sizeof(*p->excitation));
+}
+
static int g723_1_decode_frame(AVCodecContext *avctx, void *data,
int *got_frame_ptr, AVPacket *avpkt)
{
@@ -1107,14 +1310,23 @@ static int g723_1_decode_frame(AVCodecContext *avctx, void *data,
PITCH_MAX * sizeof(*p->excitation));
}
}
+ p->cng_random_seed = CNG_RANDOM_SEED;
} else {
- memset(out, 0, FRAME_LEN * 2);
- av_log(avctx, AV_LOG_WARNING,
- "G.723.1: Comfort noise generation not supported yet\n");
+ if (p->cur_frame_type == SID_FRAME) {
+ p->sid_gain = sid_gain_to_lsp_index(p->subframe[0].amp_index);
+ inverse_quant(p->sid_lsp, p->prev_lsp, p->lsp_index, 0);
+ } else if (p->past_frame_type == ACTIVE_FRAME) {
+ p->sid_gain = estimate_sid_gain(p);
+ }
- *got_frame_ptr = 1;
- *(AVFrame *)data = p->frame;
- return frame_size[dec_mode];
+ if (p->past_frame_type == ACTIVE_FRAME)
+ p->cur_gain = p->sid_gain;
+ else
+ p->cur_gain = (p->cur_gain * 7 + p->sid_gain) >> 3;
+ generate_noise(p);
+ lsp_interpolate(lpc, p->sid_lsp, p->prev_lsp);
+ /* Save the lsp_vector for the next frame */
+ memcpy(p->prev_lsp, p->sid_lsp, LPC_ORDER * sizeof(*p->prev_lsp));
}
p->past_frame_type = p->cur_frame_type;
diff --git a/libavcodec/g723_1_data.h b/libavcodec/g723_1_data.h
index 82446b3..04f8a06 100644
--- a/libavcodec/g723_1_data.h
+++ b/libavcodec/g723_1_data.h
@@ -1191,4 +1191,10 @@ static const int16_t postfilter_tbl[2][LPC_ORDER] = {
{ 24576, 18432, 13824, 10368, 7776, 5832, 4374, 3281, 2460, 1845 }
};
+static const int cng_adaptive_cb_lag[4] = { 1, 0, 1, 3 };
+
+static const int cng_filt[4] = { 273, 998, 499, 333 };
+
+static const int cng_bseg[3] = { 2048, 18432, 231233 };
+
#endif /* AVCODEC_G723_1_DATA_H */
diff --git a/tests/fate/voice.mak b/tests/fate/voice.mak
index ea765d2..b9b8aa6 100644
--- a/tests/fate/voice.mak
+++ b/tests/fate/voice.mak
@@ -27,6 +27,12 @@ fate-g723_1-dec-5: CMD = framecrc -postfilter 1 -i $(SAMPLES)/g723_1/pathd63p.tc
FATE_G723_1 += fate-g723_1-dec-6
fate-g723_1-dec-6: CMD = framecrc -postfilter 1 -i $(SAMPLES)/g723_1/tamed63p.tco
+FATE_G723_1 += fate-g723_1-dec-7
+fate-g723_1-dec-7: CMD = framecrc -postfilter 1 -i $(SAMPLES)/g723_1/dtx63b.tco
+
+FATE_G723_1 += fate-g723_1-dec-8
+fate-g723_1-dec-8: CMD = framecrc -postfilter 1 -i $(SAMPLES)/g723_1/dtx63e.tco
+
FATE_SAMPLES_AVCONV += $(FATE_G723_1)
fate-g723_1: $(FATE_G723_1)
diff --git a/tests/ref/fate/g723_1-dec-7 b/tests/ref/fate/g723_1-dec-7
new file mode 100644
index 0000000..cc30187
--- /dev/null
+++ b/tests/ref/fate/g723_1-dec-7
@@ -0,0 +1,12 @@
+#tb 0: 1/8000
+0, 0, 0, 240, 480, 0x35e4a1fd
+0, 240, 240, 240, 480, 0x2f7bdd60
+0, 480, 480, 240, 480, 0x0407e499
+0, 720, 720, 240, 480, 0x5f5ef209
+0, 960, 960, 240, 480, 0xe936e8d1
+0, 1200, 1200, 240, 480, 0x0896ddba
+0, 1440, 1440, 240, 480, 0xa885ea94
+0, 1680, 1680, 240, 480, 0x40bff3d0
+0, 1920, 1920, 240, 480, 0xe05ce4c3
+0, 2160, 2160, 240, 480, 0x80c4f790
+0, 2400, 2400, 240, 480, 0x65d5e8f9
diff --git a/tests/ref/fate/g723_1-dec-8 b/tests/ref/fate/g723_1-dec-8
new file mode 100644
index 0000000..fc4d9f3
--- /dev/null
+++ b/tests/ref/fate/g723_1-dec-8
@@ -0,0 +1,121 @@
+#tb 0: 1/8000
+0, 0, 0, 240, 480, 0x17930e0f
+0, 240, 240, 240, 480, 0x7c7f4247
+0, 480, 480, 240, 480, 0xbf3489e5
+0, 720, 720, 240, 480, 0x24319fc9
+0, 960, 960, 240, 480, 0xb327eec0
+0, 1200, 1200, 240, 480, 0xc2ddcbca
+0, 1440, 1440, 240, 480, 0xeebad740
+0, 1680, 1680, 240, 480, 0x77fcb933
+0, 1920, 1920, 240, 480, 0x9677c5b7
+0, 2160, 2160, 240, 480, 0xb49dcb9e
+0, 2400, 2400, 240, 480, 0x0e78d7e6
+0, 2640, 2640, 240, 480, 0xf752dc3e
+0, 2880, 2880, 240, 480, 0xc95af091
+0, 3120, 3120, 240, 480, 0xa25de399
+0, 3360, 3360, 240, 480, 0x34e7e0da
+0, 3600, 3600, 240, 480, 0x6c84e3f4
+0, 3840, 3840, 240, 480, 0x2c7dda20
+0, 4080, 4080, 240, 480, 0x00a5f112
+0, 4320, 4320, 240, 480, 0x943ddd89
+0, 4560, 4560, 240, 480, 0x4ad4ebac
+0, 4800, 4800, 240, 480, 0xa4ff0aa8
+0, 5040, 5040, 240, 480, 0xc0f805f2
+0, 5280, 5280, 240, 480, 0x859ce987
+0, 5520, 5520, 240, 480, 0x9ebcd0de
+0, 5760, 5760, 240, 480, 0x3de2db0b
+0, 6000, 6000, 240, 480, 0x0affea9c
+0, 6240, 6240, 240, 480, 0xcb1bf81e
+0, 6480, 6480, 240, 480, 0x8a72d71d
+0, 6720, 6720, 240, 480, 0x583cd5cd
+0, 6960, 6960, 240, 480, 0x4be7dc7b
+0, 7200, 7200, 240, 480, 0xb08108c0
+0, 7440, 7440, 240, 480, 0xd0b3ed59
+0, 7680, 7680, 240, 480, 0x7d33f822
+0, 7920, 7920, 240, 480, 0x199c0111
+0, 8160, 8160, 240, 480, 0x7d29f2a8
+0, 8400, 8400, 240, 480, 0x424dec5e
+0, 8640, 8640, 240, 480, 0x946cf258
+0, 8880, 8880, 240, 480, 0xd429da7a
+0, 9120, 9120, 240, 480, 0x0f11df46
+0, 9360, 9360, 240, 480, 0xf4dce502
+0, 9600, 9600, 240, 480, 0x01c1de78
+0, 9840, 9840, 240, 480, 0xd1d3da59
+0, 10080, 10080, 240, 480, 0x5822f3ec
+0, 10320, 10320, 240, 480, 0xadd5fe67
+0, 10560, 10560, 240, 480, 0xdcf5f2c3
+0, 10800, 10800, 240, 480, 0x5176e39b
+0, 11040, 11040, 240, 480, 0xf947e0b1
+0, 11280, 11280, 240, 480, 0x33b1eb36
+0, 11520, 11520, 240, 480, 0x57bce9bd
+0, 11760, 11760, 240, 480, 0x806eec1f
+0, 12000, 12000, 240, 480, 0x0a60f94a
+0, 12240, 12240, 240, 480, 0x9eddf27d
+0, 12480, 12480, 240, 480, 0x3d28ef2f
+0, 12720, 12720, 240, 480, 0x52f0e562
+0, 12960, 12960, 240, 480, 0xf2d6c8a0
+0, 13200, 13200, 240, 480, 0xfa0df4a1
+0, 13440, 13440, 240, 480, 0x9cccfda9
+0, 13680, 13680, 240, 480, 0xa7c1e528
+0, 13920, 13920, 240, 480, 0xe130e8f9
+0, 14160, 14160, 240, 480, 0x80f6eabe
+0, 14400, 14400, 240, 480, 0x9bbb027e
+0, 14640, 14640, 240, 480, 0x33cdea7f
+0, 14880, 14880, 240, 480, 0x84d8e761
+0, 15120, 15120, 240, 480, 0xb99ce457
+0, 15360, 15360, 240, 480, 0x5dc1e324
+0, 15600, 15600, 240, 480, 0xc914e6c3
+0, 15840, 15840, 240, 480, 0x8e77f5c2
+0, 16080, 16080, 240, 480, 0x3997034d
+0, 16320, 16320, 240, 480, 0x820cfd49
+0, 16560, 16560, 240, 480, 0x8ad5f24c
+0, 16800, 16800, 240, 480, 0xe21be71c
+0, 17040, 17040, 240, 480, 0x516ae8c8
+0, 17280, 17280, 240, 480, 0x595bdc3d
+0, 17520, 17520, 240, 480, 0x8a4bee79
+0, 17760, 17760, 240, 480, 0x307fed64
+0, 18000, 18000, 240, 480, 0xe71cf219
+0, 18240, 18240, 240, 480, 0xdb0be1a1
+0, 18480, 18480, 240, 480, 0x7947dfbd
+0, 18720, 18720, 240, 480, 0x5d90fbf0
+0, 18960, 18960, 240, 480, 0xa449fc55
+0, 19200, 19200, 240, 480, 0x45b2f979
+0, 19440, 19440, 240, 480, 0x2b2be378
+0, 19680, 19680, 240, 480, 0x2d2edf42
+0, 19920, 19920, 240, 480, 0x568ee04f
+0, 20160, 20160, 240, 480, 0x66f0debe
+0, 20400, 20400, 240, 480, 0xc943eab7
+0, 20640, 20640, 240, 480, 0xc9ade3c9
+0, 20880, 20880, 240, 480, 0x6971f92d
+0, 21120, 21120, 240, 480, 0x48d0ecbc
+0, 21360, 21360, 240, 480, 0xf641dc98
+0, 21600, 21600, 240, 480, 0xbb18e167
+0, 21840, 21840, 240, 480, 0x72ce0968
+0, 22080, 22080, 240, 480, 0x15bee6f6
+0, 22320, 22320, 240, 480, 0x93d5e91f
+0, 22560, 22560, 240, 480, 0x7aee010b
+0, 22800, 22800, 240, 480, 0x9e82dc87
+0, 23040, 23040, 240, 480, 0x4ee6f547
+0, 23280, 23280, 240, 480, 0x3072d102
+0, 23520, 23520, 240, 480, 0x74d4fb04
+0, 23760, 23760, 240, 480, 0xc670f958
+0, 24000, 24000, 240, 480, 0x3965c41f
+0, 24240, 24240, 240, 480, 0x6a2de869
+0, 24480, 24480, 240, 480, 0xa757f44b
+0, 24720, 24720, 240, 480, 0x94a5168c
+0, 24960, 24960, 240, 480, 0xef0f0c28
+0, 25200, 25200, 240, 480, 0x3770ebb3
+0, 25440, 25440, 240, 480, 0x4343de6f
+0, 25680, 25680, 240, 480, 0x3ec8d816
+0, 25920, 25920, 240, 480, 0x1661e3d3
+0, 26160, 26160, 240, 480, 0x077cd9fd
+0, 26400, 26400, 240, 480, 0xb5ece07e
+0, 26640, 26640, 240, 480, 0xf303e151
+0, 26880, 26880, 240, 480, 0x95e4d019
+0, 27120, 27120, 240, 480, 0x4bd0ddc0
+0, 27360, 27360, 240, 480, 0x6cebd341
+0, 27600, 27600, 240, 480, 0xea3fea9e
+0, 27840, 27840, 240, 480, 0x5ad30c3f
+0, 28080, 28080, 240, 480, 0x218c02a5
+0, 28320, 28320, 240, 480, 0x662decd0
+0, 28560, 28560, 240, 480, 0x6865e2f2
More information about the ffmpeg-cvslog
mailing list