Opus integration

First patch = delivery from August 22, 2012.

Review URL: https://webrtc-codereview.appspot.com/756005

git-svn-id: http://webrtc.googlecode.com/svn/trunk@2945 4adac7df-926f-26a2-2b94-8c16560cd09d
This commit is contained in:
tina.legrand@webrtc.org 2012-10-18 10:00:52 +00:00
parent 28d0140ed2
commit a7d8387bdd
40 changed files with 1098 additions and 518 deletions

3
DEPS
View File

@ -57,6 +57,9 @@ deps = {
"third_party/libyuv":
(Var("googlecode_url") % "libyuv") + "/trunk@389",
"third_party/opus/source":
"http://git.xiph.org/opus.git@v1.0.1",
"third_party/protobuf":
Var("chromium_trunk") + "/src/third_party/protobuf@" + Var("chromium_revision"),

View File

@ -85,6 +85,10 @@
# Disable the use of protocol buffers in production code.
'enable_protobuf%': 0,
# Disable Mozilla internal Opus version
'build_with_mozilla%': 0,
}, { # Settings for the standalone (not-in-Chromium) build.
'include_pulse_audio%': 1,
'include_internal_audio_device%': 1,

View File

@ -27,19 +27,27 @@
// [Voice] Codec settings
// ----------------------------------------------------------------------------
// iSAC is not included in the Mozilla build, but in all other builds.
#ifndef WEBRTC_MOZILLA_BUILD
#ifdef WEBRTC_ARCH_ARM
#define WEBRTC_CODEC_ISACFX // fix-point iSAC implementation
#define WEBRTC_CODEC_ISACFX // Fix-point iSAC implementation.
#else
#define WEBRTC_CODEC_ISAC // floating-point iSAC implementation (default)
#endif
#define WEBRTC_CODEC_ISAC // Floating-point iSAC implementation (default).
#endif // WEBRTC_ARCH_ARM
#endif // !WEBRTC_MOZILLA_BUILD
// AVT is included in all builds, along with G.711, NetEQ and CNG
// (which are mandatory and don't have any defines).
#define WEBRTC_CODEC_AVT
#ifndef WEBRTC_CHROMIUM_BUILD
// iLBC, G.722, PCM16B and Redundancy coding are excluded from Chromium and
// Mozilla builds.
#if !defined(WEBRTC_CHROMIUM_BUILD) && !defined(WEBRTC_MOZILLA_BUILD)
#define WEBRTC_CODEC_ILBC
#define WEBRTC_CODEC_G722
#define WEBRTC_CODEC_PCM16
#define WEBRTC_CODEC_RED
#endif
#endif // !WEBRTC_CHROMIUM_BUILD && !WEBRTC_MOZILLA_BUILD
// ----------------------------------------------------------------------------
// [Video] Codec settings

View File

@ -36,7 +36,7 @@ typedef struct WebRtcCngDecInst_t_ {
typedef struct WebRtcCngEncInst_t_ {
int16_t enc_nrOfCoefs;
int16_t enc_sampfreq;
uint16_t enc_sampfreq;
int16_t enc_interval;
int16_t enc_msSinceSID;
int32_t enc_Energy;

View File

@ -0,0 +1,123 @@
/*
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef WEBRTC_MODULES_AUDIO_CODING_CODECS_OPUS_INTERFACE_OPUS_INTERFACE_H_
#define WEBRTC_MODULES_AUDIO_CODING_CODECS_OPUS_INTERFACE_OPUS_INTERFACE_H_
#include "typedefs.h"
#ifdef __cplusplus
extern "C" {
#endif
// Opaque wrapper types for the codec state.
typedef struct WebRtcOpusEncInst OpusEncInst;
typedef struct WebRtcOpusDecInst OpusDecInst;
int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst, int32_t channels);
int16_t WebRtcOpus_EncoderFree(OpusEncInst* inst);
/****************************************************************************
* WebRtcOpus_Encode(...)
*
* This function encodes audio as a series of Opus frames and inserts
* it into a packet. Input buffer can be any length.
*
* Input:
* - inst : Encoder context
* - audio_in : Input speech data buffer
* - samples : Samples in audio_in
* - length_encoded_buffer : Output buffer size
*
* Output:
* - encoded : Output compressed data buffer
*
* Return value : >0 - Length (in bytes) of coded data
* -1 - Error
*/
int16_t WebRtcOpus_Encode(OpusEncInst* inst, int16_t* audio_in, int16_t samples,
int16_t length_encoded_buffer, uint8_t* encoded);
/****************************************************************************
* WebRtcOpus_SetBitRate(...)
*
* This function adjusts the target bitrate of the encoder.
*
* Input:
* - inst : Encoder context
* - rate : New target bitrate
*
* Return value : 0 - Success
* -1 - Error
*/
int16_t WebRtcOpus_SetBitRate(OpusEncInst* inst, int32_t rate);
int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, int channels);
int16_t WebRtcOpus_DecoderFree(OpusDecInst* inst);
/****************************************************************************
* WebRtcOpus_DecoderInit(...)
*
* This function resets state of the decoder.
*
* Input:
* - inst : Decoder context
*
* Return value : 0 - Success
* -1 - Error
*/
int16_t WebRtcOpus_DecoderInit(OpusDecInst* inst);
/****************************************************************************
* WebRtcOpus_Decode(...)
*
* This function decodes an Opus packet into one or more audio frames at the
* ACM interface's sampling rate (32 kHz).
*
* Input:
* - inst : Decoder context
* - encoded : Encoded data
* - encoded_bytes : Bytes in encoded vector
*
* Output:
* - decoded : The decoded vector
* - audio_type : 1 normal, 2 CNG (for Opus it should
* always return 1 since we're not using Opus's
* built-in DTX/CNG scheme)
*
* Return value : >0 - Samples in decoded vector
* -1 - Error
*/
int16_t WebRtcOpus_Decode(OpusDecInst* inst, int16_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type);
/****************************************************************************
* WebRtcOpus_DecodePlc(...)
*
* This function precesses PLC for opus frame(s).
* Input:
* - inst : Decoder context
* - number_of_lost_frames : Number of PLC frames to produce
*
* Output:
* - decoded : The decoded vector
*
* Return value : >0 - number of samples in decoded PLC vector
* -1 - Error
*/
int16_t WebRtcOpus_DecodePlc(OpusDecInst* inst, int16_t* decoded,
int16_t number_of_lost_frames);
#ifdef __cplusplus
} // extern "C"
#endif
#endif // WEBRTC_MODULES_AUDIO_CODING_CODECS_OPUS_INTERFACE_OPUS_INTERFACE_H_

View File

@ -0,0 +1,44 @@
# Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
#
# Use of this source code is governed by a BSD-style license
# that can be found in the LICENSE file in the root of the source
# tree. An additional intellectual property rights grant can be found
# in the file PATENTS. All contributing project authors may
# be found in the AUTHORS file in the root of the source tree.
{
'targets': [
{
'target_name': 'webrtc_opus',
'type': 'static_library',
'conditions': [
['build_with_mozilla==1', {
# Mozilla provides its own build of the opus library.
'include_dirs': [
'$(DIST)/include/opus',
]
}, {
'dependencies': [
'<(DEPTH)/third_party/opus/opus.gyp:opus'
],
'include_dirs': [
'<(webrtc_root)/../third_party/opus/source/include',
],
}],
],
'direct_dependent_settings': {
'conditions': [
['build_with_mozilla==1', {
'include_dirs': [
'$(DIST)/include/opus',
],
}],
],
},
'sources': [
'interface/opus_interface.h',
'opus_interface.c',
],
},
],
}

View File

@ -0,0 +1,181 @@
/*
* Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "modules/audio_coding/codecs/opus/interface/opus_interface.h"
#include <stdlib.h>
#include <string.h>
#include "opus.h"
#include "common_audio/signal_processing/resample_by_2_internal.h"
#include "common_audio/signal_processing/include/signal_processing_library.h"
enum {
/* We always produce 20ms frames. */
kWebRtcOpusMaxEncodeFrameSizeMs = 20,
/* The format allows up to 120ms frames. Since we
* don't control the other side, we must allow
* for packets that large. NetEq is currently
* limited to 60 ms on the receive side.
*/
kWebRtcOpusMaxDecodeFrameSizeMs = 120,
/* Sample count is 48 kHz * samples per frame. */
kWebRtcOpusMaxFrameSize = 48 * kWebRtcOpusMaxDecodeFrameSizeMs,
};
struct WebRtcOpusEncInst {
OpusEncoder* encoder;
};
int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst, int32_t channels) {
OpusEncInst* state;
state = (OpusEncInst*) calloc(1, sizeof(OpusEncInst));
if (state) {
int error;
state->encoder = opus_encoder_create(48000, channels, OPUS_APPLICATION_VOIP,
&error);
if (error == OPUS_OK || state->encoder != NULL ) {
*inst = state;
return 0;
}
free(state);
}
return -1;
}
int16_t WebRtcOpus_EncoderFree(OpusEncInst* inst) {
opus_encoder_destroy(inst->encoder);
return 0;
}
int16_t WebRtcOpus_Encode(OpusEncInst* inst, int16_t* audio_in, int16_t samples,
int16_t length_encoded_buffer, uint8_t* encoded) {
opus_int16* audio = (opus_int16*) audio_in;
unsigned char* coded = encoded;
int res;
if (samples > 48 * kWebRtcOpusMaxEncodeFrameSizeMs) {
return -1;
}
res = opus_encode(inst->encoder, audio, samples, coded,
length_encoded_buffer);
if (res > 0) {
return res;
}
return -1;
}
int16_t WebRtcOpus_SetBitRate(OpusEncInst* inst, int32_t rate) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_BITRATE(rate));
}
struct WebRtcOpusDecInst {
int16_t state_48_32[8];
OpusDecoder* decoder;
};
int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, int channels) {
OpusDecInst* state;
state = (OpusDecInst*) calloc(1, sizeof(OpusDecInst));
if (state) {
int error;
// Always create a 48000 Hz Opus decoder.
state->decoder = opus_decoder_create(48000, channels, &error);
if (error == OPUS_OK && state->decoder != NULL ) {
*inst = state;
return 0;
}
free(state);
state = NULL;
}
return -1;
}
int16_t WebRtcOpus_DecoderFree(OpusDecInst* inst) {
opus_decoder_destroy(inst->decoder);
free(inst);
return 0;
}
int16_t WebRtcOpus_DecoderInit(OpusDecInst* inst) {
int error = opus_decoder_ctl(inst->decoder, OPUS_RESET_STATE);
if (error == OPUS_OK) {
memset(inst->state_48_32, 0, sizeof(inst->state_48_32));
return 0;
}
return -1;
}
static int DecodeNative(OpusDecInst* inst, int16_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type) {
unsigned char* coded = (unsigned char*) encoded;
opus_int16* audio = (opus_int16*) decoded;
int res = opus_decode(inst->decoder, coded, encoded_bytes, audio,
kWebRtcOpusMaxFrameSize, 0);
/* TODO(tlegrand): set to DTX for zero-length packets? */
*audio_type = 0;
if (res > 0) {
return res;
}
return -1;
}
int16_t WebRtcOpus_Decode(OpusDecInst* inst, int16_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type) {
/* Enough for 120 ms (the largest Opus packet size) of mono audio at 48 kHz
* and resampler overlap. This will need to be enlarged for stereo decoding.
*/
int16_t buffer16[kWebRtcOpusMaxFrameSize];
int32_t buffer32[kWebRtcOpusMaxFrameSize + 7];
int decoded_samples;
int blocks;
int16_t output_samples;
int i;
/* Decode to a temporary buffer. */
decoded_samples = DecodeNative(inst, encoded, encoded_bytes, buffer16,
audio_type);
if (decoded_samples < 0) {
return -1;
}
/* Resample from 48 kHz to 32 kHz. */
for (i = 0; i < 7; i++) {
buffer32[i] = inst->state_48_32[i];
inst->state_48_32[i] = buffer16[decoded_samples -7 + i];
}
for (i = 0; i < decoded_samples; i++) {
buffer32[7 + i] = buffer16[i];
}
/* Resampling 3 samples to 2. Function divides the input in |blocks| number
* of 3-sample groups, and output is |blocks| number of 2-sample groups. */
blocks = decoded_samples / 3;
WebRtcSpl_Resample48khzTo32khz(buffer32, buffer32, blocks);
output_samples = (int16_t) (blocks * 2);
WebRtcSpl_VectorBitShiftW32ToW16(decoded, output_samples, buffer32, 15);
return output_samples;
}
int16_t WebRtcOpus_DecodePlc(OpusDecInst* inst, int16_t* decoded,
int16_t number_of_lost_frames) {
/* TODO(tlegrand): We can pass NULL to opus_decode to activate packet
* loss concealment, but I don't know how many samples
* number_of_lost_frames corresponds to. */
return -1;
}

View File

@ -81,7 +81,8 @@ WebRtc_Word32 ACMCNG::CodecDef(WebRtcNetEQ_CodecDef& codecDef,
// Then return the structure back to NetEQ to add the codec to it's
// database.
if (_sampFreqHz == 8000 || _sampFreqHz == 16000 || _sampFreqHz == 32000) {
if (_sampFreqHz == 8000 || _sampFreqHz == 16000 || _sampFreqHz == 32000 ||
_sampFreqHz == 48000) {
SET_CODEC_PAR((codecDef), kDecoderCNG, codecInst.pltype,
_decoderInstPtr, _sampFreqHz);
SET_CNG_FUNCTIONS((codecDef));

View File

@ -62,7 +62,7 @@ protected:
WebRtcCngEncInst* _encoderInstPtr;
WebRtcCngDecInst* _decoderInstPtr;
WebRtc_Word16 _sampFreqHz;
WebRtc_UWord16 _sampFreqHz;
};
} // namespace webrtc

View File

@ -86,6 +86,10 @@
#include "acm_gsmfr.h"
#include "gsmfr_interface.h"
#endif
#ifdef WEBRTC_CODEC_OPUS
#include "acm_opus.h"
#include "opus_interface.h"
#endif
#ifdef WEBRTC_CODEC_SPEEX
#include "acm_speex.h"
#include "speex_interface.h"
@ -103,22 +107,20 @@ namespace webrtc {
// codecs. Note! There are a limited number of payload types. If more codecs
// are defined they will receive reserved fixed payload types (values 69-95).
const int kDynamicPayloadtypes[ACMCodecDB::kMaxNumCodecs] = {
105, 107, 108, 109, 111, 112, 113, 114, 115, 116, 117, 120,
121, 122, 123, 124, 125, 126, 101, 100, 97, 96, 95, 94,
93, 92, 91, 90, 89, 88, 87, 86, 85, 84, 83, 82,
81, 80, 79, 78, 77, 76, 75, 74, 73, 72, 71, 70,
69,
105, 107, 108, 109, 111, 112, 113, 114, 115, 116, 117, 121,
92, 91, 90, 89, 88, 87, 86, 85, 84, 83, 82, 81,
80, 79, 78, 77, 76, 75, 74, 73, 72, 71, 70, 69,
68, 67
};
// Creates database with all supported codecs at compile time.
// Each entry needs the following parameters in the given order:
// payload type, name, sampling frequency, packet size in samples,
// number of channels, and default rate.
#if (defined(WEBRTC_CODEC_PCM16) || \
defined(WEBRTC_CODEC_AMR) || defined(WEBRTC_CODEC_AMRWB) || \
defined(WEBRTC_CODEC_CELT) || defined(WEBRTC_CODEC_G729_1) || \
defined(WEBRTC_CODEC_SPEEX) || defined(WEBRTC_CODEC_G722_1) || \
defined(WEBRTC_CODEC_G722_1C))
#if (defined(WEBRTC_CODEC_AMR) || defined(WEBRTC_CODEC_AMRWB) \
|| defined(WEBRTC_CODEC_CELT) || defined(WEBRTC_CODEC_G722_1) \
|| defined(WEBRTC_CODEC_G722_1C) || defined(WEBRTC_CODEC_G729_1) \
|| defined(WEBRTC_CODEC_PCM16) || defined(WEBRTC_CODEC_SPEEX))
static int count_database = 0;
#endif
@ -186,14 +188,19 @@ const CodecInst ACMCodecDB::database_[] = {
#ifdef WEBRTC_CODEC_GSMFR
{3, "GSM", 8000, 160, 1, 13200},
#endif
#ifdef WEBRTC_CODEC_OPUS
// Opus supports 48, 24, 16, 12, 8 kHz.
{120, "opus", 48000, 960, 1, 32000},
#endif
#ifdef WEBRTC_CODEC_SPEEX
{kDynamicPayloadtypes[count_database++], "speex", 8000, 160, 1, 11000},
{kDynamicPayloadtypes[count_database++], "speex", 16000, 320, 1, 22000},
#endif
// Comfort noise for three different sampling frequencies.
// Comfort noise for four different sampling frequencies.
{13, "CN", 8000, 240, 1, 0},
{98, "CN", 16000, 480, 1, 0},
{99, "CN", 32000, 960, 1, 0},
{100, "CN", 48000, 1440, 1, 0},
#ifdef WEBRTC_CODEC_AVT
{106, "telephone-event", 8000, 240, 1, 0},
#endif
@ -272,6 +279,11 @@ const ACMCodecDB::CodecSettings ACMCodecDB::codec_settings_[] = {
#ifdef WEBRTC_CODEC_GSMFR
{3, {160, 320, 480}, 160, 1},
#endif
#ifdef WEBRTC_CODEC_OPUS
// Opus supports frames shorter than 10ms,
// but it doesn't help us to use them.
{1, {960}, 0, 2},
#endif
#ifdef WEBRTC_CODEC_SPEEX
{3, {160, 320, 480}, 0, 1},
{3, {320, 640, 960}, 0, 1},
@ -280,6 +292,7 @@ const ACMCodecDB::CodecSettings ACMCodecDB::codec_settings_[] = {
{1, {240}, 240, 1},
{1, {480}, 480, 1},
{1, {960}, 960, 1},
{1, {1440}, 1440, 1},
#ifdef WEBRTC_CODEC_AVT
{1, {240}, 240, 1},
#endif
@ -355,6 +368,9 @@ const WebRtcNetEQDecoder ACMCodecDB::neteq_decoders_[] = {
#ifdef WEBRTC_CODEC_GSMFR
kDecoderGSMFR,
#endif
#ifdef WEBRTC_CODEC_OPUS
kDecoderOpus,
#endif
#ifdef WEBRTC_CODEC_SPEEX
kDecoderSPEEX_8,
kDecoderSPEEX_16,
@ -363,6 +379,7 @@ const WebRtcNetEQDecoder ACMCodecDB::neteq_decoders_[] = {
kDecoderCNG,
kDecoderCNG,
kDecoderCNG,
kDecoderCNG,
#ifdef WEBRTC_CODEC_AVT
kDecoderAVT,
#endif
@ -509,6 +526,9 @@ int ACMCodecDB::CodecNumber(const CodecInst* codec_inst, int* mirror_id) {
} else if (STR_CASE_CMP("g7291", codec_inst->plname) == 0) {
return IsG7291RateValid(codec_inst->rate)
? codec_id : kInvalidRate;
} else if (STR_CASE_CMP("opus", codec_inst->plname) == 0) {
return IsOpusRateValid(codec_inst->rate)
? codec_id : kInvalidRate;
} else if (STR_CASE_CMP("speex", codec_inst->plname) == 0) {
return IsSpeexRateValid(codec_inst->rate)
? codec_id : kInvalidRate;
@ -719,6 +739,10 @@ ACMGenericCodec* ACMCodecDB::CreateCodecInstance(const CodecInst* codec_inst) {
codec_id = kCNSWB;
break;
}
case 48000: {
codec_id = kCNFB;
break;
}
default: {
return NULL;
}
@ -731,6 +755,10 @@ ACMGenericCodec* ACMCodecDB::CreateCodecInstance(const CodecInst* codec_inst) {
} else if (!STR_CASE_CMP(codec_inst->plname, "G7291")) {
#ifdef WEBRTC_CODEC_G729_1
return new ACMG729_1(kG729_1);
#endif
} else if (!STR_CASE_CMP(codec_inst->plname, "opus")) {
#ifdef WEBRTC_CODEC_OPUS
return new ACMOpus(kOpus);
#endif
} else if (!STR_CASE_CMP(codec_inst->plname, "speex")) {
#ifdef WEBRTC_CODEC_SPEEX
@ -766,6 +794,10 @@ ACMGenericCodec* ACMCodecDB::CreateCodecInstance(const CodecInst* codec_inst) {
codec_id = kCNSWB;
break;
}
case 48000: {
codec_id = kCNFB;
break;
}
default: {
return NULL;
}
@ -928,6 +960,14 @@ bool ACMCodecDB::IsSpeexRateValid(int rate) {
}
}
// Checks if the bitrate is valid for Opus.
bool ACMCodecDB::IsOpusRateValid(int rate) {
if ((rate < 6000) || (rate > 510000)) {
return false;
}
return true;
}
// Checks if the bitrate is valid for Celt.
bool ACMCodecDB::IsCeltRateValid(int rate) {
if ((rate >= 48000) && (rate <= 128000)) {

View File

@ -91,6 +91,9 @@ class ACMCodecDB {
#ifdef WEBRTC_CODEC_GSMFR
, kGSMFR
#endif
#ifdef WEBRTC_CODEC_OPUS
, kOpus
#endif
#ifdef WEBRTC_CODEC_SPEEX
, kSPEEX8
, kSPEEX16
@ -98,6 +101,7 @@ class ACMCodecDB {
, kCNNB
, kCNWB
, kCNSWB
, kCNFB
#ifdef WEBRTC_CODEC_AVT
, kAVT
#endif
@ -170,6 +174,9 @@ class ACMCodecDB {
enum {kSPEEX8 = -1};
enum {kSPEEX16 = -1};
#endif
#ifndef WEBRTC_CODEC_OPUS
enum {kOpus = -1};
#endif
#ifndef WEBRTC_CODEC_AVT
enum {kAVT = -1};
#endif
@ -298,6 +305,7 @@ class ACMCodecDB {
static bool IsAMRwbRateValid(int rate);
static bool IsG7291RateValid(int rate);
static bool IsSpeexRateValid(int rate);
static bool IsOpusRateValid(int rate);
static bool IsCeltRateValid(int rate);
// Check if the payload type is valid, meaning that it is in the valid range

View File

@ -63,14 +63,15 @@ const int kIsacPacSize960 = 960;
// kPassiveDTXNB : Passive audio frame coded by narrow-band CN.
// kPassiveDTXWB : Passive audio frame coded by wide-band CN.
// kPassiveDTXSWB : Passive audio frame coded by super-wide-band CN.
//
// kPassiveDTXFB : Passive audio frame coded by full-band CN.
enum WebRtcACMEncodingType {
kNoEncoding,
kActiveNormalEncoded,
kPassiveNormalEncoded,
kPassiveDTXNB,
kPassiveDTXWB,
kPassiveDTXSWB
kPassiveDTXSWB,
kPassiveDTXFB
};
// A structure which contains codec parameters. For instance, used when

View File

@ -58,6 +58,7 @@ ACMGenericCodec::ACMGenericCodec()
_numLPCParams(kNewCNGNumPLCParams),
_sentCNPrevious(false),
_isMaster(true),
_prev_frame_cng(0),
_netEqDecodeLock(NULL),
_codecWrapperLock(*RWLockWrapper::CreateRWLock()),
_lastEncodedTimestamp(0),
@ -294,6 +295,8 @@ ACMGenericCodec::EncodeSafe(
*encodingType = kPassiveDTXWB;
} else if (sampFreqHz == 32000) {
*encodingType = kPassiveDTXSWB;
} else if (sampFreqHz == 48000) {
*encodingType = kPassiveDTXFB;
} else {
status = -1;
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
@ -1169,7 +1172,7 @@ ACMGenericCodec::EnableDTX()
}
WebRtc_UWord16 freqHz;
EncoderSampFreq(freqHz);
if(WebRtcCng_InitEnc(_ptrDTXInst, (WebRtc_Word16)freqHz,
if(WebRtcCng_InitEnc(_ptrDTXInst, freqHz,
ACM_SID_INTERVAL_MSEC, _numLPCParams) < 0)
{
// Couldn't initialize, has to return -1, and free the memory
@ -1313,6 +1316,7 @@ ACMGenericCodec::ProcessFrameVADDTX(
*samplesProcessed = 0;
return 0;
}
WebRtc_UWord16 freqHz;
EncoderSampFreq(freqHz);
@ -1321,8 +1325,8 @@ ACMGenericCodec::ProcessFrameVADDTX(
WebRtc_Word32 frameLenMsec = (((WebRtc_Word32)_frameLenSmpl * 1000) / freqHz);
WebRtc_Word16 status;
// Vector for storing maximum 30 ms of mono audio at 32 kHz
WebRtc_Word16 audio[960];
// Vector for storing maximum 30 ms of mono audio at 48 kHz.
WebRtc_Word16 audio[1440];
// Calculate number of VAD-blocks to process, and number of samples in each block.
int noSamplesToProcess[2];
@ -1378,25 +1382,33 @@ ACMGenericCodec::ProcessFrameVADDTX(
*bitStreamLenByte = 0;
for(WebRtc_Word16 n = 0; n < num10MsecFrames; n++)
{
// This block is (passive) && (vad enabled)
status = WebRtcCng_Encode(_ptrDTXInst, &audio[n*samplesIn10Msec],
samplesIn10Msec, bitStream, &bitStreamLen, 0);
// This block is (passive) && (vad enabled). If first CNG after
// speech, force SID by setting last parameter to "1".
status = WebRtcCng_Encode(_ptrDTXInst,
&audio[n*samplesIn10Msec],
samplesIn10Msec, bitStream,
&bitStreamLen, !_prev_frame_cng);
if (status < 0) {
return -1;
}
// Update previous frame was CNG.
_prev_frame_cng = 1;
*samplesProcessed += samplesIn10Msec*_noChannels;
// bitStreamLen will only be > 0 once per 100 ms
*bitStreamLenByte += bitStreamLen;
}
// Check if all samples got processed by the DTX
if(*samplesProcessed != noSamplesToProcess[i]*_noChannels) {
// Set to zero since something went wrong. Shouldn't happen.
*samplesProcessed = 0;
}
} else {
// Update previous frame was not CNG.
_prev_frame_cng = 0;
}
if(*samplesProcessed > 0)

View File

@ -1310,6 +1310,7 @@ protected:
WebRtc_UWord8 _numLPCParams;
bool _sentCNPrevious;
bool _isMaster;
int16_t _prev_frame_cng;
WebRtcACMCodecParams _encoderParams;
WebRtcACMCodecParams _decoderParams;

View File

@ -8,442 +8,256 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "acm_opus.h"
#include "acm_codec_database.h"
#include "acm_common_defs.h"
#include "acm_neteq.h"
#include "acm_opus.h"
#include "trace.h"
#include "webrtc_neteq.h"
#include "webrtc_neteq_help_macros.h"
#ifdef WEBRTC_CODEC_OPUS
// NOTE! Opus is not included in the open-source package. Modify this file or your codec
// API to match the function call and name of used Opus API file.
// #include "opus_interface.h"
#include "opus_interface.h"
#endif
namespace webrtc
{
namespace webrtc {
#ifndef WEBRTC_CODEC_OPUS
ACMOPUS::ACMOPUS(WebRtc_Word16 /* codecID */)
ACMOpus::ACMOpus(int16_t /* codecID */)
: _encoderInstPtr(NULL),
_decoderInstPtr(NULL),
_mySampFreq(0),
_myRate(0),
_opusMode(0),
_flagVBR(0) {
_sampleFreq(0),
_bitrate(0) {
return;
}
ACMOPUS::~ACMOPUS()
{
return;
ACMOpus::~ACMOpus() {
return;
}
WebRtc_Word16
ACMOPUS::InternalEncode(
WebRtc_UWord8* /* bitStream */,
WebRtc_Word16* /* bitStreamLenByte */)
{
return -1;
int16_t ACMOpus::InternalEncode(uint8_t* /* bitStream */,
int16_t* /* bitStreamLenByte */) {
return -1;
}
WebRtc_Word16
ACMOPUS::DecodeSafe(
WebRtc_UWord8* /* bitStream */,
WebRtc_Word16 /* bitStreamLenByte */,
WebRtc_Word16* /* audio */,
WebRtc_Word16* /* audioSamples */,
WebRtc_Word8* /* speechType */)
{
return -1;
int16_t ACMOpus::DecodeSafe(uint8_t* /* bitStream */,
int16_t /* bitStreamLenByte */,
int16_t* /* audio */,
int16_t* /* audioSamples */,
int8_t* /* speechType */) {
return -1;
}
WebRtc_Word16
ACMOPUS::InternalInitEncoder(
WebRtcACMCodecParams* /* codecParams */)
{
return -1;
int16_t ACMOpus::InternalInitEncoder(WebRtcACMCodecParams* /* codecParams */) {
return -1;
}
WebRtc_Word16
ACMOPUS::InternalInitDecoder(
WebRtcACMCodecParams* /* codecParams */)
{
return -1;
int16_t ACMOpus::InternalInitDecoder(WebRtcACMCodecParams* /* codecParams */) {
return -1;
}
WebRtc_Word32
ACMOPUS::CodecDef(
WebRtcNetEQ_CodecDef& /* codecDef */,
const CodecInst& /* codecInst */)
{
return -1;
int32_t ACMOpus::CodecDef(WebRtcNetEQ_CodecDef& /* codecDef */,
const CodecInst& /* codecInst */) {
return -1;
}
ACMGenericCodec*
ACMOPUS::CreateInstance(void)
{
return NULL;
ACMGenericCodec* ACMOpus::CreateInstance(void) {
return NULL;
}
WebRtc_Word16
ACMOPUS::InternalCreateEncoder()
{
return -1;
int16_t ACMOpus::InternalCreateEncoder() {
return -1;
}
void
ACMOPUS::DestructEncoderSafe()
{
return;
void ACMOpus::DestructEncoderSafe() {
return;
}
WebRtc_Word16
ACMOPUS::InternalCreateDecoder()
{
return -1;
int16_t ACMOpus::InternalCreateDecoder() {
return -1;
}
void
ACMOPUS::DestructDecoderSafe()
{
return;
void ACMOpus::DestructDecoderSafe() {
return;
}
void
ACMOPUS::InternalDestructEncoderInst(
void* /* ptrInst */)
{
return;
void ACMOpus::InternalDestructEncoderInst(void* /* ptrInst */) {
return;
}
WebRtc_Word16
ACMOPUS::SetBitRateSafe(
const WebRtc_Word32 /*rate*/ )
{
return -1;
int16_t ACMOpus::SetBitRateSafe(const int32_t /*rate*/) {
return -1;
}
#else //===================== Actual Implementation =======================
#else //===================== Actual Implementation =======================
// Remove when integrating a real Opus wrapper
extern WebRtc_Word16 WebRtcOpus_CreateEnc(OPUS_inst_t_** inst, WebRtc_Word16 samplFreq);
extern WebRtc_Word16 WebRtcOpus_CreateDec(OPUS_inst_t_** inst, WebRtc_Word16 samplFreq);
extern WebRtc_Word16 WebRtcOpus_FreeEnc(OPUS_inst_t_* inst);
extern WebRtc_Word16 WebRtcOpus_FreeDec(OPUS_inst_t_* inst);
extern WebRtc_Word16 WebRtcOpus_Encode(OPUS_inst_t_* encInst,
WebRtc_Word16* input,
WebRtc_Word16* output,
WebRtc_Word16 len,
WebRtc_Word16 byteLen);
extern WebRtc_Word16 WebRtcOpus_EncoderInit(OPUS_inst_t_* encInst,
WebRtc_Word16 samplFreq,
WebRtc_Word16 mode,
WebRtc_Word16 vbrFlag);
extern WebRtc_Word16 WebRtcOpus_Decode(OPUS_inst_t_* decInst);
extern WebRtc_Word16 WebRtcOpus_DecodeBwe(OPUS_inst_t_* decInst, WebRtc_Word16* input);
extern WebRtc_Word16 WebRtcOpus_DecodePlc(OPUS_inst_t_* decInst);
extern WebRtc_Word16 WebRtcOpus_DecoderInit(OPUS_inst_t_* decInst);
ACMOPUS::ACMOPUS(WebRtc_Word16 codecID)
ACMOpus::ACMOpus(int16_t codecID)
: _encoderInstPtr(NULL),
_decoderInstPtr(NULL),
_mySampFreq(48000), // Default sampling frequency.
_myRate(50000), // Default rate.
_opusMode(1), // Default mode is the hybrid mode.
_flagVBR(0) { // Default VBR off.
_sampleFreq(32000), // Default sampling frequency.
_bitrate(20000) { // Default bit-rate.
_codecID = codecID;
// Current implementation doesn't have DTX. That might change.
// Opus has internal DTX, but we dont use it for now.
_hasInternalDTX = false;
if (_codecID != ACMCodecDB::kOpus) {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"Wrong codec id for Opus.");
_sampleFreq = -1;
_bitrate = -1;
}
return;
}
ACMOPUS::~ACMOPUS()
{
if(_encoderInstPtr != NULL)
{
WebRtcOpus_FreeEnc(_encoderInstPtr);
_encoderInstPtr = NULL;
}
if(_decoderInstPtr != NULL)
{
WebRtcOpus_FreeDec(_decoderInstPtr);
_decoderInstPtr = NULL;
}
return;
ACMOpus::~ACMOpus() {
if (_encoderInstPtr != NULL) {
WebRtcOpus_EncoderFree(_encoderInstPtr);
_encoderInstPtr = NULL;
}
if (_decoderInstPtr != NULL) {
WebRtcOpus_DecoderFree(_decoderInstPtr);
_decoderInstPtr = NULL;
}
return;
}
WebRtc_Word16
ACMOPUS::InternalEncode(
WebRtc_UWord8* bitStream,
WebRtc_Word16* bitStreamLenByte)
{
WebRtc_Word16 noEncodedSamples = 0;
WebRtc_Word16 tmpLenByte = 0;
int16_t ACMOpus::InternalEncode(uint8_t* bitStream, int16_t* bitStreamLenByte) {
// Call Encoder.
*bitStreamLenByte = WebRtcOpus_Encode(_encoderInstPtr,
&_inAudio[_inAudioIxRead],
_frameLenSmpl,
MAX_PAYLOAD_SIZE_BYTE,
bitStream);
// Check for error reported from encoder.
if (*bitStreamLenByte < 0) {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"InternalEncode: Encode error for Opus");
*bitStreamLenByte = 0;
return -1;
}
WebRtc_Word16 byteLengthFrame = 0;
// Increment the read index. This tells the caller how far
// we have gone forward in reading the audio buffer.
_inAudioIxRead += _frameLenSmpl;
// Derive what byte-length is requested
byteLengthFrame = _myRate*_frameLenSmpl/(8*_mySampFreq);
// Call Encoder
*bitStreamLenByte = WebRtcOpus_Encode(_encoderInstPtr, &_inAudio[_inAudioIxRead],
(WebRtc_Word16*)bitStream, _frameLenSmpl, byteLengthFrame);
// increment the read index this tell the caller that how far
// we have gone forward in reading the audio buffer
_inAudioIxRead += _frameLenSmpl;
// sanity check
if(*bitStreamLenByte < 0)
{
// error has happened
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"InternalEncode: Encode error for Opus");
*bitStreamLenByte = 0;
return -1;
}
return *bitStreamLenByte;
return *bitStreamLenByte;
}
int16_t ACMOpus::DecodeSafe(uint8_t* bitStream, int16_t bitStreamLenByte,
int16_t* audio, int16_t* audioSamples,
int8_t* speechType) {
return 0;
}
int16_t ACMOpus::InternalInitEncoder(WebRtcACMCodecParams* codecParams) {
int16_t ret;
if (_encoderInstPtr != NULL) {
WebRtcOpus_EncoderFree(_encoderInstPtr);
_encoderInstPtr = NULL;
}
ret = WebRtcOpus_EncoderCreate(&_encoderInstPtr,
codecParams->codecInstant.channels);
if (ret < 0) {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"Encoder creation failed for Opus");
return ret;
}
ret = WebRtcOpus_SetBitRate(_encoderInstPtr, codecParams->codecInstant.rate);
if (ret < 0) {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"Setting initial bitrate failed for Opus");
return ret;
}
return 0;
}
WebRtc_Word16
ACMOPUS::DecodeSafe(
WebRtc_UWord8* /* bitStream */,
WebRtc_Word16 /* bitStreamLenByte */,
WebRtc_Word16* /* audio */,
WebRtc_Word16* /* audioSamples */,
WebRtc_Word8* /* speechType */)
{
int16_t ACMOpus::InternalInitDecoder(WebRtcACMCodecParams* codecParams) {
if (_decoderInstPtr != NULL) {
WebRtcOpus_DecoderFree(_decoderInstPtr);
_decoderInstPtr = NULL;
}
if (WebRtcOpus_DecoderCreate(&_decoderInstPtr,
codecParams->codecInstant.channels) < 0) {
return -1;
}
return WebRtcOpus_DecoderInit(_decoderInstPtr);
}
int32_t ACMOpus::CodecDef(WebRtcNetEQ_CodecDef& codecDef,
const CodecInst& codecInst) {
if (!_decoderInitialized) {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"CodeDef: Decoder uninitialized for Opus");
return -1;
}
// Fill up the structure by calling "SET_CODEC_PAR" & "SET_OPUS_FUNCTION."
// Then call NetEQ to add the codec to its database.
// TODO(tlegrand): Decoder is registered in NetEQ as a 32 kHz decoder, which
// is true until we have a full 48 kHz system, and remove the downsampling
// in the Opus decoder wrapper.
SET_CODEC_PAR((codecDef), kDecoderOpus, codecInst.pltype, _decoderInstPtr,
32000);
SET_OPUS_FUNCTIONS((codecDef));
return 0;
}
ACMGenericCodec* ACMOpus::CreateInstance(void) {
return NULL;
}
int16_t ACMOpus::InternalCreateEncoder() {
// Real encoder will be created in InternalInitEncoder.
return 0;
}
void ACMOpus::DestructEncoderSafe() {
if (_encoderInstPtr) {
WebRtcOpus_EncoderFree(_encoderInstPtr);
_encoderInstPtr = NULL;
}
}
int16_t ACMOpus::InternalCreateDecoder() {
// Real decoder will be created in InternalInitDecoder
return 0;
}
void ACMOpus::DestructDecoderSafe() {
_decoderInitialized = false;
if (_decoderInstPtr) {
WebRtcOpus_DecoderFree(_decoderInstPtr);
_decoderInstPtr = NULL;
}
}
void ACMOpus::InternalDestructEncoderInst(void* ptrInst) {
if (ptrInst != NULL) {
WebRtcOpus_EncoderFree((OpusEncInst*) ptrInst);
}
return;
}
int16_t ACMOpus::SetBitRateSafe(const int32_t rate) {
if (rate < 6000 || rate > 510000) {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"SetBitRateSafe: Invalid rate Opus");
return -1;
}
_bitrate = rate;
// Ask the encoder for the new rate.
if (WebRtcOpus_SetBitRate(_encoderInstPtr, _bitrate) >= 0) {
_encoderParams.codecInstant.rate = _bitrate;
return 0;
}
return -1;
}
#endif // WEBRTC_CODEC_OPUS
WebRtc_Word16
ACMOPUS::InternalInitEncoder(
WebRtcACMCodecParams* codecParams)
{
//set the bit rate and initialize
_myRate = codecParams->codecInstant.rate;
return SetBitRateSafe( (WebRtc_UWord32)_myRate);
}
WebRtc_Word16
ACMOPUS::InternalInitDecoder(
WebRtcACMCodecParams* /* codecParams */)
{
if (WebRtcOpus_DecoderInit(_decoderInstPtr) < 0)
{
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"InternalInitDecoder: init decoder failed for Opus");
return -1;
}
return 0;
}
WebRtc_Word32
ACMOPUS::CodecDef(
WebRtcNetEQ_CodecDef& codecDef,
const CodecInst& codecInst)
{
if (!_decoderInitialized)
{
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"CodeDef: Decoder uninitialized for Opus");
return -1;
}
// Fill up the structure by calling
// "SET_CODEC_PAR" & "SET_G729_FUNCTION."
// Then call NetEQ to add the codec to it's
// database.
SET_CODEC_PAR((codecDef), kDecoderOpus, codecInst.pltype,
_decoderInstPtr, 16000);
SET_OPUS_FUNCTIONS((codecDef));
return 0;
}
ACMGenericCodec*
ACMOPUS::CreateInstance(void)
{
return NULL;
}
WebRtc_Word16
ACMOPUS::InternalCreateEncoder()
{
if (WebRtcOpus_CreateEnc(&_encoderInstPtr, _mySampFreq) < 0)
{
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"InternalCreateEncoder: create encoder failed for Opus");
return -1;
}
return 0;
}
void
ACMOPUS::DestructEncoderSafe()
{
_encoderExist = false;
_encoderInitialized = false;
if(_encoderInstPtr != NULL)
{
WebRtcOpus_FreeEnc(_encoderInstPtr);
_encoderInstPtr = NULL;
}
}
WebRtc_Word16
ACMOPUS::InternalCreateDecoder()
{
if (WebRtcOpus_CreateDec(&_decoderInstPtr, _mySampFreq) < 0)
{
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"InternalCreateDecoder: create decoder failed for Opus");
return -1;
}
return 0;
}
void
ACMOPUS::DestructDecoderSafe()
{
_decoderExist = false;
_decoderInitialized = false;
if(_decoderInstPtr != NULL)
{
WebRtcOpus_FreeDec(_decoderInstPtr);
_decoderInstPtr = NULL;
}
}
void
ACMOPUS::InternalDestructEncoderInst(
void* ptrInst)
{
if(ptrInst != NULL)
{
WebRtcOpus_FreeEnc((OPUS_inst_t*)ptrInst);
}
return;
}
WebRtc_Word16
ACMOPUS::SetBitRateSafe(
const WebRtc_Word32 rate)
{
//allowed rates: {8000, 12000, 14000, 16000, 18000, 20000,
// 22000, 24000, 26000, 28000, 30000, 32000};
switch(rate)
{
case 8000:
{
_myRate = 8000;
break;
}
case 12000:
{
_myRate = 12000;
break;
}
case 14000:
{
_myRate = 14000;
break;
}
case 16000:
{
_myRate = 16000;
break;
}
case 18000:
{
_myRate = 18000;
break;
}
case 20000:
{
_myRate = 20000;
break;
}
case 22000:
{
_myRate = 22000;
break;
}
case 24000:
{
_myRate = 24000;
break;
}
case 26000:
{
_myRate = 26000;
break;
}
case 28000:
{
_myRate = 28000;
break;
}
case 30000:
{
_myRate = 30000;
break;
}
case 32000:
{
_myRate = 32000;
break;
}
default:
{
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _uniqueID,
"SetBitRateSafe: Invalid rate Opus");
return -1;
}
}
// Re-init with new rate
if (WebRtcOpus_EncoderInit(_encoderInstPtr, _mySampFreq, _opusMode, _flagVBR) >= 0)
{
_encoderParams.codecInstant.rate = _myRate;
return 0;
}
else
{
return -1;
}
}
#endif
} // namespace webrtc
} // namespace webrtc

View File

@ -12,68 +12,48 @@
#define WEBRTC_MODULES_AUDIO_CODING_MAIN_SOURCE_ACM_OPUS_H_
#include "acm_generic_codec.h"
#include "opus_interface.h"
#include "resampler.h"
// forward declaration
struct OPUS_inst_t_;
struct OPUS_inst_t_;
namespace webrtc {
namespace webrtc
{
class ACMOpus : public ACMGenericCodec {
public:
ACMOpus(int16_t codecID);
~ACMOpus();
class ACMOPUS: public ACMGenericCodec
{
public:
ACMOPUS(WebRtc_Word16 codecID);
~ACMOPUS();
// for FEC
ACMGenericCodec* CreateInstance(void);
ACMGenericCodec* CreateInstance(void);
WebRtc_Word16 InternalEncode(
WebRtc_UWord8* bitstream,
WebRtc_Word16* bitStreamLenByte);
int16_t InternalEncode(uint8_t* bitstream, int16_t* bitStreamLenByte);
WebRtc_Word16 InternalInitEncoder(
WebRtcACMCodecParams *codecParams);
int16_t InternalInitEncoder(WebRtcACMCodecParams *codecParams);
WebRtc_Word16 InternalInitDecoder(
WebRtcACMCodecParams *codecParams);
int16_t InternalInitDecoder(WebRtcACMCodecParams *codecParams);
protected:
WebRtc_Word16 DecodeSafe(
WebRtc_UWord8* bitStream,
WebRtc_Word16 bitStreamLenByte,
WebRtc_Word16* audio,
WebRtc_Word16* audioSamples,
WebRtc_Word8* speechType);
protected:
int16_t DecodeSafe(uint8_t* bitStream, int16_t bitStreamLenByte,
int16_t* audio, int16_t* audioSamples, int8_t* speechType);
WebRtc_Word32 CodecDef(
WebRtcNetEQ_CodecDef& codecDef,
const CodecInst& codecInst);
int32_t CodecDef(WebRtcNetEQ_CodecDef& codecDef, const CodecInst& codecInst);
void DestructEncoderSafe();
void DestructEncoderSafe();
void DestructDecoderSafe();
void DestructDecoderSafe();
WebRtc_Word16 InternalCreateEncoder();
int16_t InternalCreateEncoder();
WebRtc_Word16 InternalCreateDecoder();
int16_t InternalCreateDecoder();
void InternalDestructEncoderInst(
void* ptrInst);
void InternalDestructEncoderInst(void* ptrInst);
WebRtc_Word16 SetBitRateSafe(
const WebRtc_Word32 rate);
OPUS_inst_t_* _encoderInstPtr;
OPUS_inst_t_* _decoderInstPtr;
WebRtc_UWord16 _mySampFreq;
WebRtc_UWord16 _myRate;
WebRtc_Word16 _opusMode;
WebRtc_Word16 _flagVBR;
int16_t SetBitRateSafe(const int32_t rate);
OpusEncInst* _encoderInstPtr;
OpusDecInst* _decoderInstPtr;
uint16_t _sampleFreq;
uint16_t _bitrate;
};
} // namespace webrtc
} // namespace webrtc
#endif // WEBRTC_MODULES_AUDIO_CODING_MAIN_SOURCE_ACM_OPUS_H_

View File

@ -15,6 +15,7 @@
'iLBC',
'iSAC',
'iSACFix',
'webrtc_opus',
'PCM16B',
'NetEq',
'<(webrtc_root)/common_audio/common_audio.gyp:resampler',
@ -37,11 +38,13 @@
'include_dirs': [
'../interface',
'../../../interface',
'../../codecs/opus/interface',
],
'direct_dependent_settings': {
'include_dirs': [
'../interface',
'../../../interface',
'../../codecs/opus/interface',
],
},
'sources': [

View File

@ -45,6 +45,7 @@ AudioCodingModuleImpl::AudioCodingModuleImpl(const WebRtc_Word32 id)
_cng_nb_pltype(255),
_cng_wb_pltype(255),
_cng_swb_pltype(255),
_cng_fb_pltype(255),
_red_pltype(255),
_vadEnabled(false),
_dtxEnabled(false),
@ -112,6 +113,8 @@ AudioCodingModuleImpl::AudioCodingModuleImpl(const WebRtc_Word32 id)
_cng_wb_pltype = static_cast<uint8_t>(ACMCodecDB::database_[i].pltype);
} else if (ACMCodecDB::database_[i].plfreq == 32000) {
_cng_swb_pltype = static_cast<uint8_t>(ACMCodecDB::database_[i].pltype);
} else if (ACMCodecDB::database_[i].plfreq == 48000) {
_cng_fb_pltype = static_cast<uint8_t>(ACMCodecDB::database_[i].pltype);
}
}
}
@ -320,6 +323,12 @@ WebRtc_Word32 AudioCodingModuleImpl::Process() {
_isFirstRED = true;
break;
}
case kPassiveDTXFB: {
current_payload_type = _cng_fb_pltype;
frame_type = kAudioFrameCN;
_isFirstRED = true;
break;
}
}
has_data_to_send = true;
_previousPayloadType = current_payload_type;
@ -612,6 +621,10 @@ WebRtc_Word32 AudioCodingModuleImpl::RegisterSendCodec(
_cng_swb_pltype = static_cast<uint8_t>(send_codec.pltype);
break;
}
case 48000: {
_cng_fb_pltype = static_cast<uint8_t>(send_codec.pltype);
break;
}
default: {
WEBRTC_TRACE(webrtc::kTraceError, webrtc::kTraceAudioCoding, _id,
"RegisterSendCodec() failed, invalid frequency for CNG "
@ -1254,6 +1267,9 @@ WebRtc_Word32 AudioCodingModuleImpl::ReceiveFrequency() const {
CriticalSectionScoped lock(_acmCritSect);
if (DecoderParamByPlType(_lastRecvAudioCodecPlType, codec_params) < 0) {
return _netEq.CurrentSampFreqHz();
} else if (codec_params.codecInstant.plfreq == 48000) {
// TODO(tlegrand): Remove this option when we have full 48 kHz support.
return 32000;
} else {
return codec_params.codecInstant.plfreq;
}

View File

@ -279,6 +279,7 @@ class AudioCodingModuleImpl : public AudioCodingModule {
uint8_t _cng_nb_pltype;
uint8_t _cng_wb_pltype;
uint8_t _cng_swb_pltype;
uint8_t _cng_fb_pltype;
uint8_t _red_pltype;
bool _vadEnabled;
bool _dtxEnabled;

View File

@ -613,6 +613,28 @@ void TestAllCodecs::Perform() {
RegisterSendCodec('A', codec_celt, 32000, 128000, 640, 0);
Run(channel_a_to_b_);
outfile_b_.Close();
#endif
#ifdef WEBRTC_CODEC_OPUS
if (test_mode_ != 0) {
printf("===============================================================\n");
}
test_count_++;
OpenOutFile(test_count_);
char codec_opus[] = "OPUS";
RegisterSendCodec('A', codec_opus, 48000, 6000, 960, -1);
Run(channel_a_to_b_);
RegisterSendCodec('A', codec_opus, 48000, 20000, 960, -1);
Run(channel_a_to_b_);
RegisterSendCodec('A', codec_opus, 48000, 32000, 960, -1);
Run(channel_a_to_b_);
RegisterSendCodec('A', codec_opus, 48000, 48000, 960, -1);
Run(channel_a_to_b_);
RegisterSendCodec('A', codec_opus, 48000, 64000, 960, -1);
Run(channel_a_to_b_);
RegisterSendCodec('A', codec_opus, 48000, 96000, 960, -1);
Run(channel_a_to_b_);
RegisterSendCodec('A', codec_opus, 48000, 500000, 960, -1);
Run(channel_a_to_b_);
#endif
if (test_mode_ != 0) {
printf("===============================================================\n");

View File

@ -137,6 +137,21 @@ void TestVADDTX::Perform()
// Close file
_outFileB.Close();
#endif
#ifdef WEBRTC_CODEC_OPUS
// Open outputfile
OpenOutFile(testCntr++);
// Register Opus as send codec
char nameOPUS[] = "opus";
RegisterSendCodec('A', nameOPUS);
// Run the five test cased
runTestCases();
// Close file
_outFileB.Close();
#endif
if(_testMode) {
printf("Done!\n");

View File

@ -84,7 +84,7 @@ int WebRtcNetEQ_DbAdd(CodecDbInst_t *inst, enum WebRtcNetEQDecoder codec,
#ifdef NETEQ_32KHZ_WIDEBAND
&&(codec_fs!=32000)
#endif
#ifdef NETEQ_48KHZ_WIDEBAND
#if defined(NETEQ_48KHZ_WIDEBAND) || defined(NETEQ_OPUS_CODEC)
&&(codec_fs!=48000)
#endif
)
@ -114,6 +114,9 @@ int WebRtcNetEQ_DbAdd(CodecDbInst_t *inst, enum WebRtcNetEQDecoder codec,
#ifdef NETEQ_ISAC_SWB_CODEC
case kDecoderISACswb :
#endif
#ifdef NETEQ_OPUS_CODEC
case kDecoderOpus :
#endif
#ifdef NETEQ_G722_CODEC
case kDecoderG722 :
case kDecoderG722_2ch :
@ -458,6 +461,9 @@ int WebRtcNetEQ_DbGetSplitInfo(SplitInfo_t *inst, enum WebRtcNetEQDecoder codecI
#ifdef NETEQ_ISAC_SWB_CODEC
case kDecoderISACswb:
#endif
#ifdef NETEQ_OPUS_CODEC
case kDecoderOpus:
#endif
#ifdef NETEQ_ARBITRARY_CODEC
case kDecoderArbitrary:
#endif

View File

@ -62,6 +62,7 @@ enum WebRtcNetEQDecoder
kDecoderG722_1C_24,
kDecoderG722_1C_32,
kDecoderG722_1C_48,
kDecoderOpus,
kDecoderSPEEX_8,
kDecoderSPEEX_16,
kDecoderCELT_32,

View File

@ -151,7 +151,6 @@
inst.funcUpdBWEst=NULL; \
inst.funcGetErrorCode=NULL;
#define SET_PCM16B_SWB48_FUNCTIONS(inst) \
inst.funcDecode=(WebRtcNetEQ_FuncDecode)WebRtcPcm16b_DecodeW16; \
inst.funcDecodeRCU=NULL; \
@ -317,6 +316,17 @@
inst.funcUpdBWEst=NULL; \
inst.funcGetErrorCode=NULL;
#define SET_OPUS_FUNCTIONS(inst) \
inst.funcDecode=(WebRtcNetEQ_FuncDecode)WebRtcOpus_Decode; \
inst.funcDecodeRCU=NULL; \
inst.funcDecodePLC=NULL; \
inst.funcDecodeInit=(WebRtcNetEQ_FuncDecodeInit)WebRtcOpus_DecoderInit; \
inst.funcAddLatePkt=NULL; \
inst.funcGetMDinfo=NULL; \
inst.funcGetPitch=NULL; \
inst.funcUpdBWEst=NULL; \
inst.funcGetErrorCode=NULL;
#define SET_SPEEX_FUNCTIONS(inst) \
inst.funcDecode=(WebRtcNetEQ_FuncDecode)WebRtcSpeex_Decode; \
inst.funcDecodeRCU=NULL; \

View File

@ -77,6 +77,8 @@
*
* NETEQ_G722_1C_CODEC Enable G722.1 Annex C
*
* NETEQ_OPUS_CODEC Enable Opus
*
* NETEQ_SPEEX_CODEC Enable Speex (at 8 and 16 kHz sample rate)
*
* NETEQ_CELT_CODEC Enable Celt (at 32 kHz sample rate)
@ -244,6 +246,7 @@
#define NETEQ_G729_CODEC
#define NETEQ_G726_CODEC
#define NETEQ_GSMFR_CODEC
#define NETEQ_OPUS_CODEC
#define NETEQ_AMR_CODEC
#endif
@ -252,6 +255,7 @@
#define NETEQ_G722_CODEC
#define NETEQ_G722_1_CODEC
#define NETEQ_G729_1_CODEC
#define NETEQ_OPUS_CODEC
#define NETEQ_SPEEX_CODEC
#define NETEQ_AMRWB_CODEC
#define NETEQ_WIDEBAND
@ -262,6 +266,7 @@
#define NETEQ_32KHZ_WIDEBAND
#define NETEQ_G722_1C_CODEC
#define NETEQ_CELT_CODEC
#define NETEQ_OPUS_CODEC
#endif
#if (defined(NETEQ_VOICEENGINE_CODECS))
@ -295,6 +300,8 @@
#define NETEQ_G722_1C_CODEC
#define NETEQ_CELT_CODEC
/* Fullband 48 kHz codecs */
#define NETEQ_OPUS_CODEC
#endif
#if (defined(NETEQ_ALL_CODECS))
@ -331,21 +338,26 @@
/* Super wideband 48kHz codecs */
#define NETEQ_48KHZ_WIDEBAND
#define NETEQ_OPUS_CODEC
#endif
/* Max output size from decoding one frame */
#if defined(NETEQ_48KHZ_WIDEBAND)
#define NETEQ_MAX_FRAME_SIZE 2880 /* 60 ms super wideband */
#define NETEQ_MAX_OUTPUT_SIZE 3600 /* 60+15 ms super wideband (60 ms decoded + 15 ms for merge overlap) */
#define NETEQ_MAX_FRAME_SIZE 5760 /* 120 ms super wideband */
#define NETEQ_MAX_OUTPUT_SIZE 6480 /* 120+15 ms super wideband (120 ms
* decoded + 15 ms for merge overlap) */
#elif defined(NETEQ_32KHZ_WIDEBAND)
#define NETEQ_MAX_FRAME_SIZE 1920 /* 60 ms super wideband */
#define NETEQ_MAX_OUTPUT_SIZE 2400 /* 60+15 ms super wideband (60 ms decoded + 15 ms for merge overlap) */
#define NETEQ_MAX_FRAME_SIZE 3840 /* 120 ms super wideband */
#define NETEQ_MAX_OUTPUT_SIZE 4320 /* 120+15 ms super wideband (120 ms
* decoded + 15 ms for merge overlap) */
#elif defined(NETEQ_WIDEBAND)
#define NETEQ_MAX_FRAME_SIZE 960 /* 60 ms wideband */
#define NETEQ_MAX_OUTPUT_SIZE 1200 /* 60+15 ms wideband (60 ms decoded + 10 ms for merge overlap) */
#define NETEQ_MAX_FRAME_SIZE 1920 /* 120 ms wideband */
#define NETEQ_MAX_OUTPUT_SIZE 2160 /* 120+15 ms wideband (120 ms decoded +
* 15 ms for merge overlap) */
#else
#define NETEQ_MAX_FRAME_SIZE 480 /* 60 ms narrowband */
#define NETEQ_MAX_OUTPUT_SIZE 600 /* 60+15 ms narrowband (60 ms decoded + 10 ms for merge overlap) */
#define NETEQ_MAX_FRAME_SIZE 960 /* 120 ms narrowband */
#define NETEQ_MAX_OUTPUT_SIZE 1080 /* 120+15 ms narrowband (120 ms decoded
* + 15 ms for merge overlap) */
#endif

View File

@ -578,6 +578,11 @@ int WebRtcNetEQ_GetDefaultCodecSettings(const enum WebRtcNetEQDecoder *codecID,
codecBytes = 1560; /* 240ms @ 52kbps (30ms frames) */
codecBuffers = 8;
}
else if (codecID[i] == kDecoderOpus)
{
codecBytes = 15300; /* 240ms @ 510kbps (60ms frames) */
codecBuffers = 30; /* Replicating the value for PCMu/a */
}
else if ((codecID[i] == kDecoderPCM16B) ||
(codecID[i] == kDecoderPCM16B_2ch))
{

View File

@ -202,6 +202,13 @@ int WebRtcNetEQ_RecInInternal(MCUInst_t *MCU_inst, RTPPacket_t *RTPpacketInput,
/* Get CNG sample rate */
WebRtc_UWord16 fsCng = WebRtcNetEQ_DbGetSampleRate(&MCU_inst->codec_DB_inst,
RTPpacket[i_k].payloadType);
/* Force sampling frequency to 32000 Hz CNG 48000 Hz. */
/* TODO(tlegrand): remove limitation once ACM has full 48 kHz
* support. */
if (fsCng > 32000) {
fsCng = 32000;
}
if ((fsCng != MCU_inst->fs) && (fsCng > 8000))
{
/*
@ -370,10 +377,29 @@ int WebRtcNetEQ_GetTimestampScaling(MCUInst_t *MCU_inst, int rtpPayloadType)
MCU_inst->scalingFactor = kTSscalingTwo;
break;
}
case kDecoderOpus:
{
/* We resample Opus internally to 32 kHz, but timestamps
* are counted at 48 kHz. So there are two output samples
* per three RTP timestamp ticks. */
MCU_inst->scalingFactor = kTSscalingTwoThirds;
break;
}
case kDecoderAVT:
case kDecoderCNG:
{
/* do not change the timestamp scaling settings */
/* TODO(tlegrand): remove scaling once ACM has full 48 kHz
* support. */
WebRtc_UWord16 sample_freq =
WebRtcNetEQ_DbGetSampleRate(&MCU_inst->codec_DB_inst,
rtpPayloadType);
if (sample_freq == 48000) {
MCU_inst->scalingFactor = kTSscalingTwoThirds;
}
/* For sample_freq <= 32 kHz, do not change the timestamp scaling
* settings. */
break;
}
default:

View File

@ -41,8 +41,8 @@
/* Scratch usage:
Type Name size startpos endpos
WebRtc_Word16 pw16_NetEqAlgorithm_buffer 600*fs/8000 0 600*fs/8000-1
struct dspInfo 6 600*fs/8000 605*fs/8000
WebRtc_Word16 pw16_NetEqAlgorithm_buffer 1080*fs/8000 0 1080*fs/8000-1
struct dspInfo 6 1080*fs/8000 1085*fs/8000
func WebRtcNetEQ_Normal 40+495*fs/8000 0 39+495*fs/8000
func WebRtcNetEQ_Merge 40+496*fs/8000 0 39+496*fs/8000
@ -50,7 +50,7 @@
func WebRtcNetEQ_Accelerate 210 240*fs/8000 209+240*fs/8000
func WebRtcNetEQ_BGNUpdate 69 480*fs/8000 68+480*fs/8000
Total: 605*fs/8000
Total: 1086*fs/8000
*/
#define SCRATCH_ALGORITHM_BUFFER 0
@ -58,35 +58,35 @@
#define SCRATCH_NETEQ_MERGE 0
#if (defined(NETEQ_48KHZ_WIDEBAND))
#define SCRATCH_DSP_INFO 3600
#define SCRATCH_DSP_INFO 6480
#define SCRATCH_NETEQ_ACCELERATE 1440
#define SCRATCH_NETEQ_BGN_UPDATE 2880
#define SCRATCH_NETEQ_EXPAND 756
#elif (defined(NETEQ_32KHZ_WIDEBAND))
#define SCRATCH_DSP_INFO 2400
#define SCRATCH_DSP_INFO 4320
#define SCRATCH_NETEQ_ACCELERATE 960
#define SCRATCH_NETEQ_BGN_UPDATE 1920
#define SCRATCH_NETEQ_EXPAND 504
#elif (defined(NETEQ_WIDEBAND))
#define SCRATCH_DSP_INFO 1200
#define SCRATCH_DSP_INFO 2160
#define SCRATCH_NETEQ_ACCELERATE 480
#define SCRATCH_NETEQ_BGN_UPDATE 960
#define SCRATCH_NETEQ_EXPAND 252
#else /* NB */
#define SCRATCH_DSP_INFO 600
#define SCRATCH_DSP_INFO 1080
#define SCRATCH_NETEQ_ACCELERATE 240
#define SCRATCH_NETEQ_BGN_UPDATE 480
#define SCRATCH_NETEQ_EXPAND 126
#endif
#if (defined(NETEQ_48KHZ_WIDEBAND))
#define SIZE_SCRATCH_BUFFER 3636
#define SIZE_SCRATCH_BUFFER 6516
#elif (defined(NETEQ_32KHZ_WIDEBAND))
#define SIZE_SCRATCH_BUFFER 2424
#define SIZE_SCRATCH_BUFFER 4344
#elif (defined(NETEQ_WIDEBAND))
#define SIZE_SCRATCH_BUFFER 1212
#define SIZE_SCRATCH_BUFFER 2172
#else /* NB */
#define SIZE_SCRATCH_BUFFER 606
#define SIZE_SCRATCH_BUFFER 1086
#endif
#ifdef NETEQ_DELAY_LOGGING
@ -110,13 +110,15 @@ int WebRtcNetEQ_RecOutInternal(DSPInst_t *inst, WebRtc_Word16 *pw16_outData,
#ifdef SCRATCH
char pw8_ScratchBuffer[((SIZE_SCRATCH_BUFFER + 1) * 2)];
WebRtc_Word16 *pw16_scratchPtr = (WebRtc_Word16*) pw8_ScratchBuffer;
WebRtc_Word16 pw16_decoded_buffer[NETEQ_MAX_FRAME_SIZE];
/* pad with 240*fs_mult to match the overflow guard below */
WebRtc_Word16 pw16_decoded_buffer[NETEQ_MAX_FRAME_SIZE+240*6];
WebRtc_Word16 *pw16_NetEqAlgorithm_buffer = pw16_scratchPtr
+ SCRATCH_ALGORITHM_BUFFER;
DSP2MCU_info_t *dspInfo = (DSP2MCU_info_t*) (pw16_scratchPtr + SCRATCH_DSP_INFO);
#else
WebRtc_Word16 pw16_decoded_buffer[NETEQ_MAX_FRAME_SIZE];
WebRtc_Word16 pw16_NetEqAlgorithm_buffer[NETEQ_MAX_OUTPUT_SIZE];
/* pad with 240*fs_mult to match the overflow guard below */
WebRtc_Word16 pw16_decoded_buffer[NETEQ_MAX_FRAME_SIZE+240*6];
WebRtc_Word16 pw16_NetEqAlgorithm_buffer[NETEQ_MAX_OUTPUT_SIZE+240*6];
DSP2MCU_info_t dspInfoStruct;
DSP2MCU_info_t *dspInfo = &dspInfoStruct;
#endif

View File

@ -319,7 +319,13 @@ int WebRtcNetEQ_SignalMcu(MCUInst_t *inst)
WebRtc_UWord16 tempFs;
tempFs = WebRtcNetEQ_DbGetSampleRate(&inst->codec_DB_inst, payloadType);
if (tempFs > 0)
/* TODO(tlegrand): Remove this limitation once ACM has full
* 48 kHz support. */
if (tempFs > 32000)
{
inst->fs = 32000;
}
else if (tempFs > 0)
{
inst->fs = tempFs;
}

View File

@ -30,6 +30,7 @@ public:
kNbInHz = 8000,
kWbInHz = 16000,
kSwbInHz = 32000,
kFbInHz = 48000,
kLowestPossible = -1,
kDefaultFrequency = kWbInHz
};

View File

@ -282,6 +282,12 @@ WebRtc_Word32 AudioConferenceMixerImpl::Process()
SetOutputFrequency(kSwbInHz);
}
break;
case 48000:
if(OutputFrequency() != kFbInHz)
{
SetOutputFrequency(kFbInHz);
}
break;
default:
assert(false);

View File

@ -15,6 +15,7 @@
'audio_coding/codecs/ilbc/ilbc.gypi',
'audio_coding/codecs/isac/main/source/isac.gypi',
'audio_coding/codecs/isac/fix/source/isacfix.gypi',
'audio_coding/codecs/opus/opus.gypi',
'audio_coding/codecs/pcm16b/pcm16b.gypi',
'audio_coding/main/source/audio_coding_module.gypi',
'audio_coding/neteq/neteq.gypi',

View File

@ -27,6 +27,7 @@ RTPReceiverAudio::RTPReceiverAudio(const WebRtc_Word32 id):
_cngNBPayloadType(-1),
_cngWBPayloadType(-1),
_cngSWBPayloadType(-1),
_cngFBPayloadType(-1),
_cngPayloadType(-1),
_G722PayloadType(-1),
_lastReceivedG722(false),
@ -94,7 +95,7 @@ bool
RTPReceiverAudio::CNGPayloadType(const WebRtc_Word8 payloadType,
WebRtc_UWord32& frequency)
{
// we can have three CNG on 8000Hz, 16000Hz and 32000Hz
// We can have four CNG on 8000Hz, 16000Hz, 32000Hz and 48000Hz.
if(_cngNBPayloadType == payloadType)
{
frequency = 8000;
@ -129,6 +130,15 @@ RTPReceiverAudio::CNGPayloadType(const WebRtc_Word8 payloadType,
}
_cngPayloadType = _cngSWBPayloadType;
return true;
}else if(_cngFBPayloadType == payloadType)
{
frequency = 48000;
if ((_cngPayloadType != -1) &&(_cngPayloadType !=_cngFBPayloadType))
{
ResetStatistics();
}
_cngPayloadType = _cngFBPayloadType;
return true;
}else
{
// not CNG
@ -195,6 +205,8 @@ ModuleRTPUtility::Payload* RTPReceiverAudio::RegisterReceiveAudioPayload(
_cngWBPayloadType = payloadType;
} else if(frequency == 32000) {
_cngSWBPayloadType = payloadType;
} else if(frequency == 48000) {
_cngFBPayloadType = payloadType;
} else {
assert(false);
return NULL;

View File

@ -81,6 +81,7 @@ private:
WebRtc_Word8 _cngNBPayloadType;
WebRtc_Word8 _cngWBPayloadType;
WebRtc_Word8 _cngSWBPayloadType;
WebRtc_Word8 _cngFBPayloadType;
WebRtc_Word8 _cngPayloadType;
// G722 is special since it use the wrong number of RTP samples in timestamp VS. number of samples in the frame

View File

@ -38,6 +38,7 @@ RTPSenderAudio::RTPSenderAudio(const WebRtc_Word32 id, RtpRtcpClock* clock,
_cngNBPayloadType(-1),
_cngWBPayloadType(-1),
_cngSWBPayloadType(-1),
_cngFBPayloadType(-1),
_lastPayloadType(-1),
_includeAudioLevelIndication(false), // @TODO - reset at Init()?
_audioLevelIndicationID(0),
@ -101,6 +102,10 @@ WebRtc_Word32 RTPSenderAudio::RegisterAudioPayload(
} else if (frequency == 32000) {
_cngSWBPayloadType = payloadType;
} else if (frequency == 48000) {
_cngFBPayloadType = payloadType;
} else {
return -1;
}
@ -159,6 +164,15 @@ RTPSenderAudio::MarkerBit(const FrameType frameType,
return false;
}
}
if(_cngFBPayloadType != -1)
{
// we have configured SWB CNG
if(_cngFBPayloadType == payloadType)
{
// only set a marker bit when we change payload type to a non CNG
return false;
}
}
// payloadType differ
if(_lastPayloadType == -1)
{

View File

@ -117,6 +117,7 @@ private:
WebRtc_Word8 _cngNBPayloadType;
WebRtc_Word8 _cngWBPayloadType;
WebRtc_Word8 _cngSWBPayloadType;
WebRtc_Word8 _cngFBPayloadType;
WebRtc_Word8 _lastPayloadType;
// Audio level indication (https://datatracker.ietf.org/doc/draft-lennox-avt-rtp-audio-level-exthdr/)

View File

@ -6403,12 +6403,12 @@ Channel::GetPlayoutTimeStamp(WebRtc_UWord32& playoutTimestamp)
}
WebRtc_Word32 playoutFrequency = _audioCodingModule.PlayoutFrequency();
if (_audioCodingModule.ReceiveCodec(currRecCodec) == 0)
{
if (STR_CASE_CMP("G722", currRecCodec.plname) == 0)
{
playoutFrequency = 8000;
}
if (_audioCodingModule.ReceiveCodec(currRecCodec) == 0) {
if (STR_CASE_CMP("G722", currRecCodec.plname) == 0) {
playoutFrequency = 8000;
} else if (STR_CASE_CMP("opus", currRecCodec.plname) == 0) {
playoutFrequency = 48000;
}
}
timestamp -= (delayMS * (playoutFrequency/1000));
@ -6482,16 +6482,20 @@ Channel::UpdatePacketDelay(const WebRtc_UWord32 timestamp,
rtpReceiveFrequency = _audioCodingModule.ReceiveFrequency();
CodecInst currRecCodec;
if (_audioCodingModule.ReceiveCodec(currRecCodec) == 0)
{
if (STR_CASE_CMP("G722", currRecCodec.plname) == 0)
{
// Even though the actual sampling rate for G.722 audio is
// 16,000 Hz, the RTP clock rate for the G722 payload format is
// 8,000 Hz because that value was erroneously assigned in
// RFC 1890 and must remain unchanged for backward compatibility.
rtpReceiveFrequency = 8000;
}
if (_audioCodingModule.ReceiveCodec(currRecCodec) == 0) {
if (STR_CASE_CMP("G722", currRecCodec.plname) == 0) {
// Even though the actual sampling rate for G.722 audio is
// 16,000 Hz, the RTP clock rate for the G722 payload format is
// 8,000 Hz because that value was erroneously assigned in
// RFC 1890 and must remain unchanged for backward compatibility.
rtpReceiveFrequency = 8000;
} else if (STR_CASE_CMP("opus", currRecCodec.plname) == 0) {
// We are resampling Opus internally to 32,000 Hz until all our
// DSP routines can operate at 48,000 Hz, but the RTP clock
// rate for the Opus payload format is standardized to 48,000 Hz,
// because that is the maximum supported decoding sampling rate.
rtpReceiveFrequency = 48000;
}
}
const WebRtc_UWord32 timeStampDiff = timestamp - _playoutTimeStampRTP;
@ -6499,24 +6503,25 @@ Channel::UpdatePacketDelay(const WebRtc_UWord32 timestamp,
if (timeStampDiff > 0)
{
switch (rtpReceiveFrequency)
{
case 8000:
timeStampDiffMs = timeStampDiff >> 3;
break;
case 16000:
timeStampDiffMs = timeStampDiff >> 4;
break;
case 32000:
timeStampDiffMs = timeStampDiff >> 5;
break;
default:
WEBRTC_TRACE(kTraceWarning, kTraceVoice,
VoEId(_instanceId, _channelId),
"Channel::UpdatePacketDelay() invalid sample "
"rate");
timeStampDiffMs = 0;
return -1;
switch (rtpReceiveFrequency) {
case 8000:
timeStampDiffMs = static_cast<WebRtc_UWord32>(timeStampDiff >> 3);
break;
case 16000:
timeStampDiffMs = static_cast<WebRtc_UWord32>(timeStampDiff >> 4);
break;
case 32000:
timeStampDiffMs = static_cast<WebRtc_UWord32>(timeStampDiff >> 5);
break;
case 48000:
timeStampDiffMs = static_cast<WebRtc_UWord32>(timeStampDiff / 48);
break;
default:
WEBRTC_TRACE(kTraceWarning, kTraceVoice,
VoEId(_instanceId, _channelId),
"Channel::UpdatePacketDelay() invalid sample rate");
timeStampDiffMs = 0;
return -1;
}
if (timeStampDiffMs > 5000)
{
@ -6539,20 +6544,23 @@ Channel::UpdatePacketDelay(const WebRtc_UWord32 timestamp,
if (sequenceNumber - _previousSequenceNumber == 1)
{
WebRtc_UWord16 packetDelayMs = 0;
switch (rtpReceiveFrequency)
{
case 8000:
packetDelayMs = (WebRtc_UWord16)(
switch (rtpReceiveFrequency) {
case 8000:
packetDelayMs = static_cast<WebRtc_UWord16>(
(timestamp - _previousTimestamp) >> 3);
break;
case 16000:
packetDelayMs = (WebRtc_UWord16)(
case 16000:
packetDelayMs = static_cast<WebRtc_UWord16>(
(timestamp - _previousTimestamp) >> 4);
break;
case 32000:
packetDelayMs = (WebRtc_UWord16)(
case 32000:
packetDelayMs = static_cast<WebRtc_UWord16>(
(timestamp - _previousTimestamp) >> 5);
break;
case 48000:
packetDelayMs = static_cast<WebRtc_UWord16>(
(timestamp - _previousTimestamp) / 48);
break;
}
if (packetDelayMs >= 10 && packetDelayMs <= 60)

View File

@ -322,8 +322,14 @@ void TransmitMixer::CheckForSendCodecChanges() {
if (codec.channels == 2)
stereo_codec_ = true;
if (codec.plfreq > _mixingFrequency)
// TODO(tlegrand): Remove once we have full 48 kHz support in
// Audio Coding Module.
if (codec.plfreq > 32000) {
_mixingFrequency = 32000;
} else if (codec.plfreq > _mixingFrequency) {
_mixingFrequency = codec.plfreq;
}
}
channel = sc.GetNextChannel(iterator);
}

184
third_party/opus/opus.gyp vendored Normal file
View File

@ -0,0 +1,184 @@
# Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
#
# Use of this source code is governed by a BSD-style license
# that can be found in the LICENSE file in the root of the source
# tree. An additional intellectual property rights grant can be found
# in the file PATENTS. All contributing project authors may
# be found in the AUTHORS file in the root of the source tree.
{
'targets': [
{
'target_name': 'opus',
'type': 'static_library',
'defines': [
'OPUS_BUILD',
],
'conditions': [
['OS=="linux"', {
'cflags': [
'-std=c99',
],
}],
['OS != "win"', {
'defines': [
'HAVE_LRINTF',
'VAR_ARRAYS',
],
}],
['OS == "win"', {
'defines': [
'USE_ALLOCA',
['inline','__inline'],
],
'msvs_disabled_warnings':[
4305, # Disable truncation warning in /source/celt/pitch.c[line 80]
],
}],
],
'include_dirs': [
'source/celt',
'source/include',
'source/silk',
'source/silk/float',
'source/src',
],
'sources': [
# opus wrapper/glue
'source/src/opus.c',
'source/src/opus_decoder.c',
'source/src/opus_encoder.c',
'source/src/repacketizer.c',
# celt sub-codec
'source/celt/bands.c',
'source/celt/celt.c',
'source/celt/celt_lpc.c',
'source/celt/cwrs.c',
'source/celt/entcode.c',
'source/celt/entdec.c',
'source/celt/entenc.c',
'source/celt/kiss_fft.c',
'source/celt/laplace.c',
'source/celt/mathops.c',
'source/celt/mdct.c',
'source/celt/modes.c',
'source/celt/pitch.c',
'source/celt/quant_bands.c',
'source/celt/rate.c',
'source/celt/vq.c',
# silk sub-codec
'source/silk/A2NLSF.c',
'source/silk/ana_filt_bank_1.c',
'source/silk/biquad_alt.c',
'source/silk/bwexpander.c',
'source/silk/bwexpander_32.c',
'source/silk/check_control_input.c',
'source/silk/CNG.c',
'source/silk/code_signs.c',
'source/silk/control_audio_bandwidth.c',
'source/silk/control_codec.c',
'source/silk/control_SNR.c',
'source/silk/debug.c',
'source/silk/decode_core.c',
'source/silk/decode_frame.c',
'source/silk/decode_indices.c',
'source/silk/decode_parameters.c',
'source/silk/decode_pitch.c',
'source/silk/decode_pulses.c',
'source/silk/decoder_set_fs.c',
'source/silk/dec_API.c',
'source/silk/enc_API.c',
'source/silk/encode_indices.c',
'source/silk/encode_pulses.c',
'source/silk/gain_quant.c',
'source/silk/HP_variable_cutoff.c',
'source/silk/init_decoder.c',
'source/silk/init_encoder.c',
'source/silk/inner_prod_aligned.c',
'source/silk/interpolate.c',
'source/silk/lin2log.c',
'source/silk/log2lin.c',
'source/silk/LPC_analysis_filter.c',
'source/silk/LPC_inv_pred_gain.c',
'source/silk/LP_variable_cutoff.c',
'source/silk/NLSF2A.c',
'source/silk/NLSF_decode.c',
'source/silk/NLSF_encode.c',
'source/silk/NLSF_del_dec_quant.c',
'source/silk/NLSF_stabilize.c',
'source/silk/NLSF_unpack.c',
'source/silk/NLSF_VQ.c',
'source/silk/NLSF_VQ_weights_laroia.c',
'source/silk/NSQ.c',
'source/silk/NSQ_del_dec.c',
'source/silk/pitch_est_tables.c',
'source/silk/PLC.c',
'source/silk/process_NLSFs.c',
'source/silk/quant_LTP_gains.c',
'source/silk/resampler.c',
'source/silk/resampler_down2.c',
'source/silk/resampler_down2_3.c',
'source/silk/resampler_private_AR2.c',
'source/silk/resampler_private_down_FIR.c',
'source/silk/resampler_private_IIR_FIR.c',
'source/silk/resampler_private_up2_HQ.c',
'source/silk/resampler_rom.c',
'source/silk/shell_coder.c',
'source/silk/sigm_Q15.c',
'source/silk/sort.c',
'source/silk/stereo_decode_pred.c',
'source/silk/stereo_encode_pred.c',
'source/silk/stereo_find_predictor.c',
'source/silk/stereo_LR_to_MS.c',
'source/silk/stereo_MS_to_LR.c',
'source/silk/stereo_quant_pred.c',
'source/silk/sum_sqr_shift.c',
'source/silk/table_LSF_cos.c',
'source/silk/tables_gain.c',
'source/silk/tables_LTP.c',
'source/silk/tables_NLSF_CB_NB_MB.c',
'source/silk/tables_NLSF_CB_WB.c',
'source/silk/tables_other.c',
'source/silk/tables_pitch_lag.c',
'source/silk/tables_pulses_per_block.c',
'source/silk/VAD.c',
'source/silk/VQ_WMat_EC.c',
# silk floating point engine
'source/silk/float/apply_sine_window_FLP.c',
'source/silk/float/autocorrelation_FLP.c',
'source/silk/float/burg_modified_FLP.c',
'source/silk/float/bwexpander_FLP.c',
'source/silk/float/corrMatrix_FLP.c',
'source/silk/float/encode_frame_FLP.c',
'source/silk/float/energy_FLP.c',
'source/silk/float/find_LPC_FLP.c',
'source/silk/float/find_LTP_FLP.c',
'source/silk/float/find_pitch_lags_FLP.c',
'source/silk/float/find_pred_coefs_FLP.c',
'source/silk/float/inner_product_FLP.c',
'source/silk/float/k2a_FLP.c',
'source/silk/float/levinsondurbin_FLP.c',
'source/silk/float/LPC_analysis_filter_FLP.c',
'source/silk/float/LPC_inv_pred_gain_FLP.c',
'source/silk/float/LTP_analysis_filter_FLP.c',
'source/silk/float/LTP_scale_ctrl_FLP.c',
'source/silk/float/noise_shape_analysis_FLP.c',
'source/silk/float/pitch_analysis_core_FLP.c',
'source/silk/float/prefilter_FLP.c',
'source/silk/float/process_gains_FLP.c',
'source/silk/float/regularize_correlations_FLP.c',
'source/silk/float/residual_energy_FLP.c',
'source/silk/float/scale_copy_vector_FLP.c',
'source/silk/float/scale_vector_FLP.c',
'source/silk/float/schur_FLP.c',
'source/silk/float/solve_LS_FLP.c',
'source/silk/float/sort_FLP.c',
'source/silk/float/warped_autocorrelation_FLP.c',
'source/silk/float/wrappers_FLP.c',
]
}
]
}

View File

@ -15,6 +15,7 @@ supplement_gypi = """#!/usr/bin/env python
{
'variables': {
'build_with_chromium': 0,
'build_with_mozilla': 0,
}
}
"""