Opus multistream.

This is a backwards-compatible change. It makes WebRTC use the Opus
multistream decoder for all Opus packets. Single-stream packets are a
special case of multistream ones (with stream=1).

The tricky parts are 'WebRtcOpus_GetMaxPlaybackRate' and
'WebRtcOpus_GetSurroundParameters'. GetMaxPlaybackRate is supposed to
do what opus_encoder_ctl(encoder, OPUS_GET_MAX_BANDWIDTH(&bandwidth))
did when we had single-stream encoders. Now there may be several
independent encoders with possibly different BANDWIDTH. The new
GetMaxPlaybackRate queries all of them, and returns a playback rate if
all the encoder's rates are equal.

WebRtcOpus_GetSurroundParameters is a configuration convention. It
maps the number of channels to a multi-stream encoder/decoder
configuration. As described in RFC 7845
https://tools.ietf.org/html/rfc7845#section-5.1.1, a multi-stream
encoder/decoder needs a number of streams, number of coupled streams
and a 255-byte mapping array. The function GetSurroundParameters
computes all of these from the number of channels. [1, 2, 4, 6, 8]
channels are supported.

Bug: webrtc:8649
Change-Id: I271de8e387d738254d6aa53af7fcf8644a53edb5
Reviewed-on: https://webrtc-review.googlesource.com/c/111750
Commit-Queue: Alex Loiko <aleloi@webrtc.org>
Reviewed-by: Minyue Li <minyue@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#26293}
This commit is contained in:
Alex Loiko
2019-01-17 12:32:11 +01:00
committed by Commit Bot
parent fe626f53f0
commit 83ed89a45f
7 changed files with 199 additions and 54 deletions

View File

@ -37,6 +37,40 @@ enum {
kWebRtcOpusDefaultFrameSize = 960,
};
int16_t GetSurroundParameters(int channels,
int *streams,
int *coupled_streams,
unsigned char *mapping) {
int opus_error;
int ret = 0;
// Use 'surround encoder create' to get values for 'coupled_streams',
// 'streams' and 'mapping'.
OpusMSEncoder* ms_encoder_ptr = opus_multistream_surround_encoder_create(
48000,
channels,
/* mapping family */ channels <= 2 ? 0 : 1,
streams,
coupled_streams,
mapping,
OPUS_APPLICATION_VOIP, // Application type shouldn't affect
// streams/mapping values.
&opus_error);
// This shouldn't fail; if it fails,
// signal an error and return invalid values.
if (opus_error != OPUS_OK || ms_encoder_ptr == NULL) {
ret = -1;
*streams = -1;
*coupled_streams = -1;
}
// We don't need the encoder.
if (ms_encoder_ptr != NULL) {
opus_multistream_encoder_destroy(ms_encoder_ptr);
}
return ret;
}
int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst,
size_t channels,
int32_t application) {
@ -55,12 +89,26 @@ int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst,
return -1;
}
unsigned char mapping[255];
memset(mapping, 0, 255);
int streams = -1;
int coupled_streams = -1;
OpusEncInst* state = calloc(1, sizeof(OpusEncInst));
RTC_DCHECK(state);
int error;
state->encoder = opus_encoder_create(48000, (int)channels, opus_app,
&error);
state->encoder = opus_multistream_surround_encoder_create(
48000,
channels,
/* mapping family */ channels <= 2 ? 0 : 1,
&streams,
&coupled_streams,
mapping,
opus_app,
&error);
if (error != OPUS_OK || !state->encoder) {
WebRtcOpus_EncoderFree(state);
return -1;
@ -75,7 +123,7 @@ int16_t WebRtcOpus_EncoderCreate(OpusEncInst** inst,
int16_t WebRtcOpus_EncoderFree(OpusEncInst* inst) {
if (inst) {
opus_encoder_destroy(inst->encoder);
opus_multistream_encoder_destroy(inst->encoder);
free(inst);
return 0;
} else {
@ -94,11 +142,11 @@ int WebRtcOpus_Encode(OpusEncInst* inst,
return -1;
}
res = opus_encode(inst->encoder,
(const opus_int16*)audio_in,
(int)samples,
encoded,
(opus_int32)length_encoded_buffer);
res = opus_multistream_encode(inst->encoder,
(const opus_int16*)audio_in,
(int)samples,
encoded,
(opus_int32)length_encoded_buffer);
if (res <= 0) {
return -1;
@ -122,7 +170,7 @@ int WebRtcOpus_Encode(OpusEncInst* inst,
int16_t WebRtcOpus_SetBitRate(OpusEncInst* inst, int32_t rate) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_BITRATE(rate));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_BITRATE(rate));
} else {
return -1;
}
@ -130,8 +178,8 @@ int16_t WebRtcOpus_SetBitRate(OpusEncInst* inst, int32_t rate) {
int16_t WebRtcOpus_SetPacketLossRate(OpusEncInst* inst, int32_t loss_rate) {
if (inst) {
return opus_encoder_ctl(inst->encoder,
OPUS_SET_PACKET_LOSS_PERC(loss_rate));
return opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_PACKET_LOSS_PERC(loss_rate));
} else {
return -1;
}
@ -154,13 +202,46 @@ int16_t WebRtcOpus_SetMaxPlaybackRate(OpusEncInst* inst, int32_t frequency_hz) {
} else {
set_bandwidth = OPUS_BANDWIDTH_FULLBAND;
}
return opus_encoder_ctl(inst->encoder,
OPUS_SET_MAX_BANDWIDTH(set_bandwidth));
return opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_MAX_BANDWIDTH(set_bandwidth));
}
int16_t WebRtcOpus_GetMaxPlaybackRate(OpusEncInst* const inst,
int32_t* result_hz) {
opus_int32 max_bandwidth;
int s;
int ret;
max_bandwidth = 0;
ret = OPUS_OK;
s = 0;
while (ret == OPUS_OK) {
OpusEncoder *enc;
opus_int32 bandwidth;
ret = opus_multistream_encoder_ctl(
inst->encoder,
OPUS_MULTISTREAM_GET_ENCODER_STATE(s, &enc));
if (ret == OPUS_BAD_ARG)
break;
if (ret != OPUS_OK)
return -1;
if (opus_encoder_ctl(enc, OPUS_GET_MAX_BANDWIDTH(&bandwidth)) != OPUS_OK)
return -1;
if (max_bandwidth != 0 && max_bandwidth != bandwidth)
return -1;
max_bandwidth = bandwidth;
s++;
}
*result_hz = max_bandwidth;
return 0;
}
int16_t WebRtcOpus_EnableFec(OpusEncInst* inst) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_INBAND_FEC(1));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_INBAND_FEC(1));
} else {
return -1;
}
@ -168,7 +249,7 @@ int16_t WebRtcOpus_EnableFec(OpusEncInst* inst) {
int16_t WebRtcOpus_DisableFec(OpusEncInst* inst) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_INBAND_FEC(0));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_INBAND_FEC(0));
} else {
return -1;
}
@ -184,21 +265,21 @@ int16_t WebRtcOpus_EnableDtx(OpusEncInst* inst) {
// last long during a pure silence, if the signal type is not forced.
// TODO(minyue): Remove the signal type forcing when Opus DTX works properly
// without it.
int ret = opus_encoder_ctl(inst->encoder,
OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE));
int ret = opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_SIGNAL(OPUS_SIGNAL_VOICE));
if (ret != OPUS_OK)
return ret;
return opus_encoder_ctl(inst->encoder, OPUS_SET_DTX(1));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_DTX(1));
}
int16_t WebRtcOpus_DisableDtx(OpusEncInst* inst) {
if (inst) {
int ret = opus_encoder_ctl(inst->encoder,
OPUS_SET_SIGNAL(OPUS_AUTO));
int ret = opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_SIGNAL(OPUS_AUTO));
if (ret != OPUS_OK)
return ret;
return opus_encoder_ctl(inst->encoder, OPUS_SET_DTX(0));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_DTX(0));
} else {
return -1;
}
@ -206,7 +287,7 @@ int16_t WebRtcOpus_DisableDtx(OpusEncInst* inst) {
int16_t WebRtcOpus_EnableCbr(OpusEncInst* inst) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_VBR(0));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_VBR(0));
} else {
return -1;
}
@ -214,7 +295,7 @@ int16_t WebRtcOpus_EnableCbr(OpusEncInst* inst) {
int16_t WebRtcOpus_DisableCbr(OpusEncInst* inst) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_VBR(1));
return opus_multistream_encoder_ctl(inst->encoder, OPUS_SET_VBR(1));
} else {
return -1;
}
@ -222,7 +303,8 @@ int16_t WebRtcOpus_DisableCbr(OpusEncInst* inst) {
int16_t WebRtcOpus_SetComplexity(OpusEncInst* inst, int32_t complexity) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_COMPLEXITY(complexity));
return opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_COMPLEXITY(complexity));
} else {
return -1;
}
@ -233,7 +315,8 @@ int32_t WebRtcOpus_GetBandwidth(OpusEncInst* inst) {
return -1;
}
int32_t bandwidth;
if (opus_encoder_ctl(inst->encoder, OPUS_GET_BANDWIDTH(&bandwidth)) == 0) {
if (opus_multistream_encoder_ctl(inst->encoder,
OPUS_GET_BANDWIDTH(&bandwidth)) == 0) {
return bandwidth;
} else {
return -1;
@ -243,7 +326,8 @@ int32_t WebRtcOpus_GetBandwidth(OpusEncInst* inst) {
int16_t WebRtcOpus_SetBandwidth(OpusEncInst* inst, int32_t bandwidth) {
if (inst) {
return opus_encoder_ctl(inst->encoder, OPUS_SET_BANDWIDTH(bandwidth));
return opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_BANDWIDTH(bandwidth));
} else {
return -1;
}
@ -253,10 +337,10 @@ int16_t WebRtcOpus_SetForceChannels(OpusEncInst* inst, size_t num_channels) {
if (!inst)
return -1;
if (num_channels == 0) {
return opus_encoder_ctl(inst->encoder,
return opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_FORCE_CHANNELS(OPUS_AUTO));
} else if (num_channels == 1 || num_channels == 2) {
return opus_encoder_ctl(inst->encoder,
return opus_multistream_encoder_ctl(inst->encoder,
OPUS_SET_FORCE_CHANNELS(num_channels));
} else {
return -1;
@ -268,16 +352,31 @@ int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, size_t channels) {
OpusDecInst* state;
if (inst != NULL) {
/* Create Opus decoder state. */
// Create Opus decoder state.
state = (OpusDecInst*) calloc(1, sizeof(OpusDecInst));
if (state == NULL) {
return -1;
}
/* Create new memory, always at 48000 Hz. */
state->decoder = opus_decoder_create(48000, (int)channels, &error);
unsigned char mapping[255];
memset(mapping, 0, 255);
int streams = -1;
int coupled_streams = -1;
if (GetSurroundParameters(channels, &streams,
&coupled_streams, mapping) != 0) {
free(state);
return -1;
}
// Create new memory, always at 48000 Hz.
state->decoder = opus_multistream_decoder_create(
48000, (int)channels,
/* streams = */ streams,
/* coupled streams = */ coupled_streams,
mapping,
&error);
if (error == OPUS_OK && state->decoder != NULL) {
/* Creation of memory all ok. */
// Creation of memory all ok.
state->channels = channels;
state->prev_decoded_samples = kWebRtcOpusDefaultFrameSize;
state->in_dtx_mode = 0;
@ -285,9 +384,9 @@ int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, size_t channels) {
return 0;
}
/* If memory allocation was unsuccessful, free the entire state. */
// If memory allocation was unsuccessful, free the entire state.
if (state->decoder) {
opus_decoder_destroy(state->decoder);
opus_multistream_decoder_destroy(state->decoder);
}
free(state);
}
@ -296,7 +395,7 @@ int16_t WebRtcOpus_DecoderCreate(OpusDecInst** inst, size_t channels) {
int16_t WebRtcOpus_DecoderFree(OpusDecInst* inst) {
if (inst) {
opus_decoder_destroy(inst->decoder);
opus_multistream_decoder_destroy(inst->decoder);
free(inst);
return 0;
} else {
@ -309,7 +408,7 @@ size_t WebRtcOpus_DecoderChannels(OpusDecInst* inst) {
}
void WebRtcOpus_DecoderInit(OpusDecInst* inst) {
opus_decoder_ctl(inst->decoder, OPUS_RESET_STATE);
opus_multistream_decoder_ctl(inst->decoder, OPUS_RESET_STATE);
inst->in_dtx_mode = 0;
}
@ -324,6 +423,10 @@ static int16_t DetermineAudioType(OpusDecInst* inst, size_t encoded_bytes) {
// fact a 1-byte TOC with a 1-byte payload. That will be erroneously
// interpreted as comfort noise output, but such a payload is probably
// faulty anyway.
// TODO(webrtc:10218): This is wrong for multistream opus. Then are several
// single-stream packets glued together with some packet size bytes in
// between. See https://tools.ietf.org/html/rfc6716#appendix-B
inst->in_dtx_mode = 1;
return 2; // Comfort noise.
} else {
@ -338,8 +441,9 @@ static int16_t DetermineAudioType(OpusDecInst* inst, size_t encoded_bytes) {
static int DecodeNative(OpusDecInst* inst, const uint8_t* encoded,
size_t encoded_bytes, int frame_size,
int16_t* decoded, int16_t* audio_type, int decode_fec) {
int res = opus_decode(inst->decoder, encoded, (opus_int32)encoded_bytes,
(opus_int16*)decoded, frame_size, decode_fec);
int res = opus_multistream_decode(
inst->decoder, encoded, (opus_int32)encoded_bytes,
(opus_int16*)decoded, frame_size, decode_fec);
if (res <= 0)
return -1;