zqz/platform-external-webrtc
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Revert 6458 "Since NetEq4 is ready to handle 48 kHz codec, it is..."

Browse Source 
> Since NetEq4 is ready to handle 48 kHz codec, it is good to remove the 48-to-32kHz downsampling of Opus output. This facilitates webrtc to make full use of Opus's bandwidth and eliminates unneeded computation in resampling.
> 
> TEST=passed_all_trybots
> R=henrik.lundin@webrtc.org, tina.legrand@webrtc.org
> 
> Review URL: https://webrtc-codereview.appspot.com/16619005

TBR=minyue@webrtc.org

Review URL: https://webrtc-codereview.appspot.com/17719004

git-svn-id: http://webrtc.googlecode.com/svn/trunk@6462 4adac7df-926f-26a2-2b94-8c16560cd09d
This commit is contained in:
minyue@webrtc.org
2014-06-17 09:50:12 +00:00
parent 851a09e71a
commit d42da54768
12 changed files with 433 additions and 249 deletions
Download Patch File Download Diff File Expand all files Collapse all files

259
webrtc/modules/audio_coding/codecs/opus/opus_interface.c
View File

@ -15,6 +15,9 @@
#include "opus.h"
#include "webrtc/common_audio/signal_processing/resample_by_2_internal.h"
#include "webrtc/common_audio/signal_processing/include/signal_processing_library.h"
enum {
/* Maximum supported frame size in WebRTC is 60 ms. */
kWebRtcOpusMaxEncodeFrameSizeMs = 60,
@ -28,6 +31,17 @@ enum {
* milliseconds. */
kWebRtcOpusMaxFrameSizePerChannel = 48 * kWebRtcOpusMaxDecodeFrameSizeMs,
/* Maximum sample count per frame is 48 kHz * maximum frame size in
* milliseconds * maximum number of channels. */
kWebRtcOpusMaxFrameSize = kWebRtcOpusMaxFrameSizePerChannel * 2,
/* Maximum sample count per channel for output resampled to 32 kHz,
* 32 kHz * maximum frame size in milliseconds. */
kWebRtcOpusMaxFrameSizePerChannel32kHz = 32 * kWebRtcOpusMaxDecodeFrameSizeMs,
/* Number of samples in resampler state. */
kWebRtcOpusStateSize = 7,
/* Default frame size, 20 ms @ 48 kHz, in samples (for one channel). */
kWebRtcOpusDefaultFrameSize = 960,
};
@ -129,6 +143,8 @@ int16_t WebRtcOpus_SetComplexity(OpusEncInst* inst, int32_t complexity) {
}
struct WebRtcOpusDecInst {
int16_t state_48_32_left[8];
int16_t state_48_32_right[8];
OpusDecoder* decoder_left;
OpusDecoder* decoder_right;
int prev_decoded_samples;
@ -189,6 +205,8 @@ int WebRtcOpus_DecoderChannels(OpusDecInst* inst) {
int16_t WebRtcOpus_DecoderInitNew(OpusDecInst* inst) {
int error = opus_decoder_ctl(inst->decoder_left, OPUS_RESET_STATE);
if (error == OPUS_OK) {
memset(inst->state_48_32_left, 0, sizeof(inst->state_48_32_left));
memset(inst->state_48_32_right, 0, sizeof(inst->state_48_32_right));
return 0;
}
return -1;
@ -197,6 +215,7 @@ int16_t WebRtcOpus_DecoderInitNew(OpusDecInst* inst) {
int16_t WebRtcOpus_DecoderInit(OpusDecInst* inst) {
int error = opus_decoder_ctl(inst->decoder_left, OPUS_RESET_STATE);
if (error == OPUS_OK) {
memset(inst->state_48_32_left, 0, sizeof(inst->state_48_32_left));
return 0;
}
return -1;
@ -205,6 +224,7 @@ int16_t WebRtcOpus_DecoderInit(OpusDecInst* inst) {
int16_t WebRtcOpus_DecoderInitSlave(OpusDecInst* inst) {
int error = opus_decoder_ctl(inst->decoder_right, OPUS_RESET_STATE);
if (error == OPUS_OK) {
memset(inst->state_48_32_right, 0, sizeof(inst->state_48_32_right));
return 0;
}
return -1;
@ -247,29 +267,124 @@ static int DecodeFec(OpusDecoder* inst, const int16_t* encoded,
return -1;
}
/* Resample from 48 to 32 kHz. Length of state is assumed to be
* kWebRtcOpusStateSize (7).
*/
static int WebRtcOpus_Resample48to32(const int16_t* samples_in, int length,
int16_t* state, int16_t* samples_out) {
int i;
int blocks;
int16_t output_samples;
int32_t buffer32[kWebRtcOpusMaxFrameSizePerChannel + kWebRtcOpusStateSize];
/* Resample from 48 kHz to 32 kHz. */
for (i = 0; i < kWebRtcOpusStateSize; i++) {
buffer32[i] = state[i];
state[i] = samples_in[length - kWebRtcOpusStateSize + i];
}
for (i = 0; i < length; i++) {
buffer32[kWebRtcOpusStateSize + i] = samples_in[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.
* When this is removed, the compensation in WebRtcOpus_DurationEst should be
* removed too. */
blocks = length / 3;
WebRtcSpl_Resample48khzTo32khz(buffer32, buffer32, blocks);
output_samples = (int16_t) (blocks * 2);
WebRtcSpl_VectorBitShiftW32ToW16(samples_out, output_samples, buffer32, 15);
return output_samples;
}
static int WebRtcOpus_DeInterleaveResample(OpusDecInst* inst, int16_t* input,
int sample_pairs, int16_t* output) {
int i;
int16_t buffer_left[kWebRtcOpusMaxFrameSizePerChannel];
int16_t buffer_right[kWebRtcOpusMaxFrameSizePerChannel];
int16_t buffer_out[kWebRtcOpusMaxFrameSizePerChannel32kHz];
int resampled_samples;
/* De-interleave the signal in left and right channel. */
for (i = 0; i < sample_pairs; i++) {
/* Take every second sample, starting at the first sample. */
buffer_left[i] = input[i * 2];
buffer_right[i] = input[i * 2 + 1];
}
/* Resample from 48 kHz to 32 kHz for left channel. */
resampled_samples = WebRtcOpus_Resample48to32(
buffer_left, sample_pairs, inst->state_48_32_left, buffer_out);
/* Add samples interleaved to output vector. */
for (i = 0; i < resampled_samples; i++) {
output[i * 2] = buffer_out[i];
}
/* Resample from 48 kHz to 32 kHz for right channel. */
resampled_samples = WebRtcOpus_Resample48to32(
buffer_right, sample_pairs, inst->state_48_32_right, buffer_out);
/* Add samples interleaved to output vector. */
for (i = 0; i < resampled_samples; i++) {
output[i * 2 + 1] = buffer_out[i];
}
return resampled_samples;
}
int16_t WebRtcOpus_DecodeNew(OpusDecInst* inst, const uint8_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type) {
/* |buffer| is big enough for 120 ms (the largest Opus packet size) of stereo
* audio at 48 kHz. */
int16_t buffer[kWebRtcOpusMaxFrameSize];
int16_t* coded = (int16_t*)encoded;
int decoded_samples;
int resampled_samples;
/* If mono case, just do a regular call to the decoder.
* If stereo, we need to de-interleave the stereo output into blocks with
* left and right channel. Each block is resampled to 32 kHz, and then
* interleaved again. */
/* Decode to a temporary buffer. */
decoded_samples = DecodeNative(inst->decoder_left, coded, encoded_bytes,
kWebRtcOpusMaxFrameSizePerChannel,
decoded, audio_type);
buffer, audio_type);
if (decoded_samples < 0) {
return -1;
}
if (inst->channels == 2) {
/* De-interleave and resample. */
resampled_samples = WebRtcOpus_DeInterleaveResample(inst,
buffer,
decoded_samples,
decoded);
} else {
/* Resample from 48 kHz to 32 kHz. Filter state memory for left channel is
* used for mono signals. */
resampled_samples = WebRtcOpus_Resample48to32(buffer,
decoded_samples,
inst->state_48_32_left,
decoded);
}
/* Update decoded sample memory, to be used by the PLC in case of losses. */
inst->prev_decoded_samples = decoded_samples;
return decoded_samples;
return resampled_samples;
}
int16_t WebRtcOpus_Decode(OpusDecInst* inst, const int16_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type) {
/* |buffer16| is big enough for 120 ms (the largestOpus packet size) of
* stereo audio at 48 kHz. */
int16_t buffer16[kWebRtcOpusMaxFrameSize];
int decoded_samples;
int16_t output_samples;
int i;
/* If mono case, just do a regular call to the decoder.
@ -278,82 +393,120 @@ int16_t WebRtcOpus_Decode(OpusDecInst* inst, const int16_t* encoded,
* This is to make stereo work with the current setup of NetEQ, which
* requires two calls to the decoder to produce stereo. */
/* Decode to a temporary buffer. */
decoded_samples = DecodeNative(inst->decoder_left, encoded, encoded_bytes,
kWebRtcOpusMaxFrameSizePerChannel, decoded,
kWebRtcOpusMaxFrameSizePerChannel, buffer16,
audio_type);
if (decoded_samples < 0) {
return -1;
}
if (inst->channels == 2) {
/* The parameter |decoded_samples| holds the number of samples pairs, in
* case of stereo. Number of samples in |decoded| equals |decoded_samples|
* case of stereo. Number of samples in |buffer16| equals |decoded_samples|
* times 2. */
for (i = 0; i < decoded_samples; i++) {
/* Take every second sample, starting at the first sample. This gives
* the left channel. */
decoded[i] = decoded[i * 2];
buffer16[i] = buffer16[i * 2];
}
}
/* Resample from 48 kHz to 32 kHz. */
output_samples = WebRtcOpus_Resample48to32(buffer16, decoded_samples,
inst->state_48_32_left, decoded);
/* Update decoded sample memory, to be used by the PLC in case of losses. */
inst->prev_decoded_samples = decoded_samples;
return decoded_samples;
return output_samples;
}
int16_t WebRtcOpus_DecodeSlave(OpusDecInst* inst, const int16_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type) {
/* |buffer16| is big enough for 120 ms (the largestOpus packet size) of
* stereo audio at 48 kHz. */
int16_t buffer16[kWebRtcOpusMaxFrameSize];
int decoded_samples;
int16_t output_samples;
int i;
/* Decode to a temporary buffer. */
decoded_samples = DecodeNative(inst->decoder_right, encoded, encoded_bytes,
kWebRtcOpusMaxFrameSizePerChannel, decoded,
kWebRtcOpusMaxFrameSizePerChannel, buffer16,
audio_type);
if (decoded_samples < 0) {
return -1;
}
if (inst->channels == 2) {
/* The parameter |decoded_samples| holds the number of samples pairs, in
* case of stereo. Number of samples in |decoded| equals |decoded_samples|
* case of stereo. Number of samples in |buffer16| equals |decoded_samples|
* times 2. */
for (i = 0; i < decoded_samples; i++) {
/* Take every second sample, starting at the second sample. This gives
* the right channel. */
decoded[i] = decoded[i * 2 + 1];
buffer16[i] = buffer16[i * 2 + 1];
}
} else {
/* Decode slave should never be called for mono packets. */
return -1;
}
/* Resample from 48 kHz to 32 kHz. */
output_samples = WebRtcOpus_Resample48to32(buffer16, decoded_samples,
inst->state_48_32_right, decoded);
return decoded_samples;
return output_samples;
}
int16_t WebRtcOpus_DecodePlc(OpusDecInst* inst, int16_t* decoded,
int16_t number_of_lost_frames) {
int16_t buffer[kWebRtcOpusMaxFrameSize];
int16_t audio_type = 0;
int decoded_samples;
int resampled_samples;
int plc_samples;
/* The number of samples we ask for is |number_of_lost_frames| times
* |prev_decoded_samples_|. Limit the number of samples to maximum
* |kWebRtcOpusMaxFrameSizePerChannel|. */
/* If mono case, just do a regular call to the plc function, before
* resampling.
* If stereo, we need to de-interleave the stereo output into blocks with
* left and right channel. Each block is resampled to 32 kHz, and then
* interleaved again. */
/* Decode to a temporary buffer. The number of samples we ask for is
* |number_of_lost_frames| times |prev_decoded_samples_|. Limit the number
* of samples to maximum |kWebRtcOpusMaxFrameSizePerChannel|. */
plc_samples = number_of_lost_frames * inst->prev_decoded_samples;
plc_samples = (plc_samples <= kWebRtcOpusMaxFrameSizePerChannel) ?
plc_samples : kWebRtcOpusMaxFrameSizePerChannel;
decoded_samples = DecodeNative(inst->decoder_left, NULL, 0, plc_samples,
decoded, &audio_type);
buffer, &audio_type);
if (decoded_samples < 0) {
return -1;
}
return decoded_samples;
if (inst->channels == 2) {
/* De-interleave and resample. */
resampled_samples = WebRtcOpus_DeInterleaveResample(inst,
buffer,
decoded_samples,
decoded);
} else {
/* Resample from 48 kHz to 32 kHz. Filter state memory for left channel is
* used for mono signals. */
resampled_samples = WebRtcOpus_Resample48to32(buffer,
decoded_samples,
inst->state_48_32_left,
decoded);
}
return resampled_samples;
}
int16_t WebRtcOpus_DecodePlcMaster(OpusDecInst* inst, int16_t* decoded,
int16_t number_of_lost_frames) {
int16_t buffer[kWebRtcOpusMaxFrameSize];
int decoded_samples;
int resampled_samples;
int16_t audio_type = 0;
int plc_samples;
int i;
@ -364,35 +517,42 @@ int16_t WebRtcOpus_DecodePlcMaster(OpusDecInst* inst, int16_t* decoded,
* output. This is to make stereo work with the current setup of NetEQ, which
* requires two calls to the decoder to produce stereo. */
/* The number of samples we ask for is |number_of_lost_frames| times
* |prev_decoded_samples_|. Limit the number of samples to maximum
* |kWebRtcOpusMaxFrameSizePerChannel|. */
/* Decode to a temporary buffer. The number of samples we ask for is
* |number_of_lost_frames| times |prev_decoded_samples_|. Limit the number
* of samples to maximum |kWebRtcOpusMaxFrameSizePerChannel|. */
plc_samples = number_of_lost_frames * inst->prev_decoded_samples;
plc_samples = (plc_samples <= kWebRtcOpusMaxFrameSizePerChannel) ?
plc_samples : kWebRtcOpusMaxFrameSizePerChannel;
decoded_samples = DecodeNative(inst->decoder_left, NULL, 0, plc_samples,
decoded, &audio_type);
buffer, &audio_type);
if (decoded_samples < 0) {
return -1;
}
if (inst->channels == 2) {
/* The parameter |decoded_samples| holds the number of sample pairs, in
* case of stereo. The original number of samples in |decoded| equals
* case of stereo. The original number of samples in |buffer| equals
* |decoded_samples| times 2. */
for (i = 0; i < decoded_samples; i++) {
/* Take every second sample, starting at the first sample. This gives
* the left channel. */
decoded[i] = decoded[i * 2];
buffer[i] = buffer[i * 2];
}
}
return decoded_samples;
/* Resample from 48 kHz to 32 kHz for left channel. */
resampled_samples = WebRtcOpus_Resample48to32(buffer,
decoded_samples,
inst->state_48_32_left,
decoded);
return resampled_samples;
}
int16_t WebRtcOpus_DecodePlcSlave(OpusDecInst* inst, int16_t* decoded,
int16_t number_of_lost_frames) {
int16_t buffer[kWebRtcOpusMaxFrameSize];
int decoded_samples;
int resampled_samples;
int16_t audio_type = 0;
int plc_samples;
int i;
@ -403,35 +563,44 @@ int16_t WebRtcOpus_DecodePlcSlave(OpusDecInst* inst, int16_t* decoded,
return -1;
}
/* The number of samples we ask for is |number_of_lost_frames| times
* |prev_decoded_samples_|. Limit the number of samples to maximum
* |kWebRtcOpusMaxFrameSizePerChannel|. */
/* Decode to a temporary buffer. The number of samples we ask for is
* |number_of_lost_frames| times |prev_decoded_samples_|. Limit the number
* of samples to maximum |kWebRtcOpusMaxFrameSizePerChannel|. */
plc_samples = number_of_lost_frames * inst->prev_decoded_samples;
plc_samples = (plc_samples <= kWebRtcOpusMaxFrameSizePerChannel)
? plc_samples : kWebRtcOpusMaxFrameSizePerChannel;
decoded_samples = DecodeNative(inst->decoder_right, NULL, 0, plc_samples,
decoded, &audio_type);
buffer, &audio_type);
if (decoded_samples < 0) {
return -1;
}
/* The parameter |decoded_samples| holds the number of sample pairs,
* The original number of samples in |decoded| equals |decoded_samples|
* The original number of samples in |buffer| equals |decoded_samples|
* times 2. */
for (i = 0; i < decoded_samples; i++) {
/* Take every second sample, starting at the second sample. This gives
* the right channel. */
decoded[i] = decoded[i * 2 + 1];
buffer[i] = buffer[i * 2 + 1];
}
return decoded_samples;
/* Resample from 48 kHz to 32 kHz for left channel. */
resampled_samples = WebRtcOpus_Resample48to32(buffer,
decoded_samples,
inst->state_48_32_right,
decoded);
return resampled_samples;
}
int16_t WebRtcOpus_DecodeFec(OpusDecInst* inst, const uint8_t* encoded,
int16_t encoded_bytes, int16_t* decoded,
int16_t* audio_type) {
/* |buffer| is big enough for 120 ms (the largest Opus packet size) of stereo
* audio at 48 kHz. */
int16_t buffer[kWebRtcOpusMaxFrameSize];
int16_t* coded = (int16_t*)encoded;
int decoded_samples;
int resampled_samples;
int fec_samples;
if (WebRtcOpus_PacketHasFec(encoded, encoded_bytes) != 1) {
@ -440,13 +609,33 @@ int16_t WebRtcOpus_DecodeFec(OpusDecInst* inst, const uint8_t* encoded,
fec_samples = opus_packet_get_samples_per_frame(encoded, 48000);
/* Decode to a temporary buffer. */
decoded_samples = DecodeFec(inst->decoder_left, coded, encoded_bytes,
fec_samples, decoded, audio_type);
fec_samples, buffer, audio_type);
if (decoded_samples < 0) {
return -1;
}
return decoded_samples;
/* If mono case, just do a regular call to the decoder.
* If stereo, we need to de-interleave the stereo output into blocks with
* left and right channel. Each block is resampled to 32 kHz, and then
* interleaved again. */
if (inst->channels == 2) {
/* De-interleave and resample. */
resampled_samples = WebRtcOpus_DeInterleaveResample(inst,
buffer,
decoded_samples,
decoded);
} else {
/* Resample from 48 kHz to 32 kHz. Filter state memory for left channel is
* used for mono signals. */
resampled_samples = WebRtcOpus_Resample48to32(buffer,
decoded_samples,
inst->state_48_32_left,
decoded);
}
return resampled_samples;
}
int WebRtcOpus_DurationEst(OpusDecInst* inst,
@ -463,6 +652,10 @@ int WebRtcOpus_DurationEst(OpusDecInst* inst,
/* Invalid payload duration. */
return 0;
}
/* Compensate for the down-sampling from 48 kHz to 32 kHz.
* This should be removed when the resampling in WebRtcOpus_Decode is
* removed. */
samples = samples * 2 / 3;
return samples;
}
@ -478,6 +671,10 @@ int WebRtcOpus_FecDurationEst(const uint8_t* payload,
/* Invalid payload duration. */
return 0;
}
/* Compensate for the down-sampling from 48 kHz to 32 kHz.
* This should be removed when the resampling in WebRtcOpus_Decode is
* removed. */
samples = samples * 2 / 3;
return samples;
}