mirror of https://github.com/FFmpeg/FFmpeg.git
This isn't too useful as a normal codec, but can be used in voip style applications. The decoder updates the noise generator parameters when a packet is given to it for decoding, but if called with an empty packet, it generates more noise according to the last parameters. Signed-off-by: Martin Storsjö <martin@martin.st>pull/6/merge
Martin Storsjö
12 years ago
parent
8b25a20efb
commit
9b500b8f6c
8 changed files with 291 additions and 1 deletions
@ -0,0 +1,162 @@ |
||||
/*
|
||||
* RFC 3389 comfort noise generator |
||||
* Copyright (c) 2012 Martin Storsjo |
||||
* |
||||
* This file is part of Libav. |
||||
* |
||||
* Libav is free software; you can redistribute it and/or |
||||
* modify it under the terms of the GNU Lesser General Public |
||||
* License as published by the Free Software Foundation; either |
||||
* version 2.1 of the License, or (at your option) any later version. |
||||
* |
||||
* Libav is distributed in the hope that it will be useful, |
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
||||
* Lesser General Public License for more details. |
||||
* |
||||
* You should have received a copy of the GNU Lesser General Public |
||||
* License along with Libav; if not, write to the Free Software |
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
||||
*/ |
||||
|
||||
#include <math.h> |
||||
|
||||
#include "libavutil/common.h" |
||||
#include "avcodec.h" |
||||
#include "celp_filters.h" |
||||
#include "libavutil/lfg.h" |
||||
|
||||
typedef struct CNGContext { |
||||
AVFrame avframe; |
||||
float *refl_coef, *target_refl_coef; |
||||
float *lpc_coef; |
||||
int order; |
||||
int energy, target_energy; |
||||
float *filter_out; |
||||
float *excitation; |
||||
AVLFG lfg; |
||||
} CNGContext; |
||||
|
||||
static av_cold int cng_decode_close(AVCodecContext *avctx) |
||||
{ |
||||
CNGContext *p = avctx->priv_data; |
||||
av_free(p->refl_coef); |
||||
av_free(p->target_refl_coef); |
||||
av_free(p->lpc_coef); |
||||
av_free(p->filter_out); |
||||
av_free(p->excitation); |
||||
return 0; |
||||
} |
||||
|
||||
static av_cold int cng_decode_init(AVCodecContext *avctx) |
||||
{ |
||||
CNGContext *p = avctx->priv_data; |
||||
|
||||
avctx->sample_fmt = AV_SAMPLE_FMT_S16; |
||||
avctx->channels = 1; |
||||
avctx->sample_rate = 8000; |
||||
|
||||
avcodec_get_frame_defaults(&p->avframe); |
||||
avctx->coded_frame = &p->avframe; |
||||
p->order = 12; |
||||
avctx->frame_size = 640; |
||||
p->refl_coef = av_mallocz(p->order * sizeof(*p->refl_coef)); |
||||
p->target_refl_coef = av_mallocz(p->order * sizeof(*p->target_refl_coef)); |
||||
p->lpc_coef = av_mallocz(p->order * sizeof(*p->lpc_coef)); |
||||
p->filter_out = av_mallocz((avctx->frame_size + p->order) * |
||||
sizeof(*p->filter_out)); |
||||
p->excitation = av_mallocz(avctx->frame_size * sizeof(*p->excitation)); |
||||
if (!p->refl_coef || !p->target_refl_coef || !p->lpc_coef || |
||||
!p->filter_out || !p->excitation) { |
||||
cng_decode_close(avctx); |
||||
return AVERROR(ENOMEM); |
||||
} |
||||
|
||||
av_lfg_init(&p->lfg, 0); |
||||
|
||||
return 0; |
||||
} |
||||
|
||||
static void make_lpc_coefs(float *lpc, const float *refl, int order) |
||||
{ |
||||
float buf[100]; |
||||
float *next, *cur; |
||||
int m, i; |
||||
next = buf; |
||||
cur = lpc; |
||||
for (m = 0; m < order; m++) { |
||||
next[m] = refl[m]; |
||||
for (i = 0; i < m; i++) |
||||
next[i] = cur[i] + refl[m] * cur[m - i - 1]; |
||||
FFSWAP(float*, next, cur); |
||||
} |
||||
if (cur != lpc) |
||||
memcpy(lpc, cur, sizeof(*lpc) * order); |
||||
} |
||||
|
||||
static int cng_decode_frame(AVCodecContext *avctx, void *data, |
||||
int *got_frame_ptr, AVPacket *avpkt) |
||||
{ |
||||
|
||||
CNGContext *p = avctx->priv_data; |
||||
int buf_size = avpkt->size; |
||||
int ret, i; |
||||
int16_t *buf_out; |
||||
float e = 1.0; |
||||
float scaling; |
||||
|
||||
if (avpkt->size) { |
||||
float dbov = -avpkt->data[0] / 10.0; |
||||
p->target_energy = 1081109975 * pow(10, dbov) * 0.75; |
||||
memset(p->target_refl_coef, 0, sizeof(p->refl_coef)); |
||||
for (i = 0; i < FFMIN(avpkt->size - 1, p->order); i++) { |
||||
p->target_refl_coef[i] = (avpkt->data[1 + i] - 127) / 128.0; |
||||
} |
||||
make_lpc_coefs(p->lpc_coef, p->refl_coef, p->order); |
||||
} |
||||
|
||||
p->energy = p->energy / 2 + p->target_energy / 2; |
||||
for (i = 0; i < p->order; i++) |
||||
p->refl_coef[i] = 0.6 *p->refl_coef[i] + 0.4 * p->target_refl_coef[i]; |
||||
|
||||
for (i = 0; i < p->order; i++) |
||||
e *= 1.0 - p->refl_coef[i]*p->refl_coef[i]; |
||||
|
||||
scaling = sqrt(e * p->energy / 1081109975); |
||||
for (i = 0; i < avctx->frame_size; i++) { |
||||
int r = (av_lfg_get(&p->lfg) & 0xffff) - 0x8000; |
||||
p->excitation[i] = scaling * r; |
||||
} |
||||
ff_celp_lp_synthesis_filterf(p->filter_out + p->order, p->lpc_coef, |
||||
p->excitation, avctx->frame_size, p->order); |
||||
|
||||
p->avframe.nb_samples = avctx->frame_size; |
||||
if ((ret = avctx->get_buffer(avctx, &p->avframe)) < 0) { |
||||
av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); |
||||
return ret; |
||||
} |
||||
buf_out = (int16_t *)p->avframe.data[0]; |
||||
for (i = 0; i < avctx->frame_size; i++) |
||||
buf_out[i] = p->filter_out[i + p->order]; |
||||
memcpy(p->filter_out, p->filter_out + avctx->frame_size, |
||||
p->order * sizeof(*p->filter_out)); |
||||
|
||||
*got_frame_ptr = 1; |
||||
*(AVFrame *)data = p->avframe; |
||||
|
||||
return buf_size; |
||||
} |
||||
|
||||
AVCodec ff_comfortnoise_decoder = { |
||||
.name = "comfortnoise", |
||||
.type = AVMEDIA_TYPE_AUDIO, |
||||
.id = AV_CODEC_ID_COMFORT_NOISE, |
||||
.priv_data_size = sizeof(CNGContext), |
||||
.init = cng_decode_init, |
||||
.decode = cng_decode_frame, |
||||
.close = cng_decode_close, |
||||
.long_name = NULL_IF_CONFIG_SMALL("RFC 3389 comfort noise generator"), |
||||
.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16, |
||||
AV_SAMPLE_FMT_NONE }, |
||||
.capabilities = CODEC_CAP_DELAY | CODEC_CAP_DR1, |
||||
}; |
||||
@ -0,0 +1,116 @@ |
||||
/*
|
||||
* RFC 3389 comfort noise generator |
||||
* Copyright (c) 2012 Martin Storsjo |
||||
* |
||||
* This file is part of Libav. |
||||
* |
||||
* Libav is free software; you can redistribute it and/or |
||||
* modify it under the terms of the GNU Lesser General Public |
||||
* License as published by the Free Software Foundation; either |
||||
* version 2.1 of the License, or (at your option) any later version. |
||||
* |
||||
* Libav is distributed in the hope that it will be useful, |
||||
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
||||
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
||||
* Lesser General Public License for more details. |
||||
* |
||||
* You should have received a copy of the GNU Lesser General Public |
||||
* License along with Libav; if not, write to the Free Software |
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
||||
*/ |
||||
|
||||
#include <math.h> |
||||
|
||||
#include "libavutil/common.h" |
||||
#include "avcodec.h" |
||||
#include "internal.h" |
||||
#include "lpc.h" |
||||
|
||||
typedef struct CNGContext { |
||||
LPCContext lpc; |
||||
int order; |
||||
int32_t *samples32; |
||||
double *ref_coef; |
||||
} CNGContext; |
||||
|
||||
static av_cold int cng_encode_close(AVCodecContext *avctx) |
||||
{ |
||||
CNGContext *p = avctx->priv_data; |
||||
ff_lpc_end(&p->lpc); |
||||
av_free(p->samples32); |
||||
av_free(p->ref_coef); |
||||
return 0; |
||||
} |
||||
|
||||
static av_cold int cng_encode_init(AVCodecContext *avctx) |
||||
{ |
||||
CNGContext *p = avctx->priv_data; |
||||
int ret; |
||||
|
||||
if (avctx->channels != 1) { |
||||
av_log(avctx, AV_LOG_ERROR, "Only mono supported\n"); |
||||
return AVERROR(EINVAL); |
||||
} |
||||
|
||||
avctx->frame_size = 640; |
||||
p->order = 10; |
||||
if ((ret = ff_lpc_init(&p->lpc, avctx->frame_size, p->order, FF_LPC_TYPE_LEVINSON)) < 0) |
||||
return ret; |
||||
p->samples32 = av_malloc(avctx->frame_size * sizeof(*p->samples32)); |
||||
p->ref_coef = av_malloc(p->order * sizeof(*p->ref_coef)); |
||||
if (!p->samples32 || !p->ref_coef) { |
||||
cng_encode_close(avctx); |
||||
return AVERROR(ENOMEM); |
||||
} |
||||
|
||||
return 0; |
||||
} |
||||
|
||||
static int cng_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, |
||||
const AVFrame *frame, int *got_packet_ptr) |
||||
{ |
||||
CNGContext *p = avctx->priv_data; |
||||
int ret, i; |
||||
double energy = 0; |
||||
int qdbov; |
||||
int16_t *samples = (int16_t*) frame->data[0]; |
||||
|
||||
if ((ret = ff_alloc_packet(avpkt, 1 + p->order))) { |
||||
av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n"); |
||||
return ret; |
||||
} |
||||
|
||||
for (i = 0; i < frame->nb_samples; i++) { |
||||
p->samples32[i] = samples[i]; |
||||
energy += samples[i] * samples[i]; |
||||
} |
||||
energy /= frame->nb_samples; |
||||
if (energy > 0) { |
||||
double dbov = 10 * log10(energy / 1081109975); |
||||
qdbov = av_clip(-floor(dbov), 0, 127); |
||||
} else { |
||||
qdbov = 127; |
||||
} |
||||
ret = ff_lpc_calc_ref_coefs(&p->lpc, p->samples32, p->order, p->ref_coef); |
||||
avpkt->data[0] = qdbov; |
||||
for (i = 0; i < p->order; i++) |
||||
avpkt->data[1 + i] = p->ref_coef[i] * 127 + 127; |
||||
|
||||
*got_packet_ptr = 1; |
||||
avpkt->size = 1 + p->order; |
||||
|
||||
return 0; |
||||
} |
||||
|
||||
AVCodec ff_comfortnoise_encoder = { |
||||
.name = "comfortnoise", |
||||
.type = AVMEDIA_TYPE_AUDIO, |
||||
.id = AV_CODEC_ID_COMFORT_NOISE, |
||||
.priv_data_size = sizeof(CNGContext), |
||||
.init = cng_encode_init, |
||||
.encode2 = cng_encode_frame, |
||||
.close = cng_encode_close, |
||||
.long_name = NULL_IF_CONFIG_SMALL("RFC 3389 comfort noise generator"), |
||||
.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16, |
||||
AV_SAMPLE_FMT_NONE }, |
||||
}; |
||||
Loading…
Reference in new issue