Make Opus PLC always output 10ms audio.
BUG: b/143582588
Change-Id: I41ad5f4f91d9af3f595666a8f32b7ab5382605bd
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/158672
Commit-Queue: Minyue Li <minyue@webrtc.org>
Reviewed-by: Jakob Ivarsson <jakobi@webrtc.org>
Reviewed-by: Henrik Lundin <henrik.lundin@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29733}
diff --git a/modules/audio_coding/BUILD.gn b/modules/audio_coding/BUILD.gn
index 6f49cfe..909bc75 100644
--- a/modules/audio_coding/BUILD.gn
+++ b/modules/audio_coding/BUILD.gn
@@ -840,6 +840,7 @@
"../../rtc_base:checks",
"../../rtc_base:ignore_wundef",
"../../rtc_base:rtc_base_approved",
+ "../../system_wrappers:field_trial",
]
}
diff --git a/modules/audio_coding/codecs/opus/opus_fec_test.cc b/modules/audio_coding/codecs/opus/opus_fec_test.cc
index 47e40c6..1923647 100644
--- a/modules/audio_coding/codecs/opus/opus_fec_test.cc
+++ b/modules/audio_coding/codecs/opus/opus_fec_test.cc
@@ -154,8 +154,13 @@
WebRtcOpus_DecodeFec(opus_decoder_, &bit_stream_[0], encoded_bytes_,
&out_data_[0], &audio_type);
} else {
- value_1 =
- WebRtcOpus_Decode(opus_decoder_, NULL, 0, &out_data_[0], &audio_type);
+ // Call decoder PLC.
+ while (value_1 < static_cast<int>(block_length_sample_)) {
+ int ret = WebRtcOpus_Decode(opus_decoder_, NULL, 0, &out_data_[value_1],
+ &audio_type);
+ EXPECT_EQ(ret, sampling_khz_ * 10); // Should return 10 ms of samples.
+ value_1 += ret;
+ }
}
EXPECT_EQ(static_cast<int>(block_length_sample_), value_1);
}
diff --git a/modules/audio_coding/codecs/opus/opus_inst.h b/modules/audio_coding/codecs/opus/opus_inst.h
index 9c3acb3b..148baa2 100644
--- a/modules/audio_coding/codecs/opus/opus_inst.h
+++ b/modules/audio_coding/codecs/opus/opus_inst.h
@@ -31,6 +31,7 @@
OpusDecoder* decoder;
OpusMSDecoder* multistream_decoder;
int prev_decoded_samples;
+ bool plc_use_prev_decoded_samples;
size_t channels;
int in_dtx_mode;
int sample_rate_hz;
diff --git a/modules/audio_coding/codecs/opus/opus_interface.cc b/modules/audio_coding/codecs/opus/opus_interface.cc
index fc3d3ff..2f475cb 100644
--- a/modules/audio_coding/codecs/opus/opus_interface.cc
+++ b/modules/audio_coding/codecs/opus/opus_interface.cc
@@ -11,6 +11,7 @@
#include "modules/audio_coding/codecs/opus/opus_interface.h"
#include "rtc_base/checks.h"
+#include "system_wrappers/include/field_trial.h"
enum {
#if WEBRTC_OPUS_SUPPORT_120MS_PTIME
@@ -25,8 +26,14 @@
* side, we must allow for packets of that size. NetEq is currently limited
* to 60 ms on the receive side. */
kWebRtcOpusMaxDecodeFrameSizeMs = 120,
+
+ // Duration of audio that each call to packet loss concealment covers.
+ kWebRtcOpusPlcFrameSizeMs = 10,
};
+constexpr char kPlcUsePrevDecodedSamplesFieldTrial[] =
+ "WebRTC-Audio-OpusPlcUsePrevDecodedSamples";
+
static int FrameSizePerChannel(int frame_size_ms, int sample_rate_hz) {
RTC_DCHECK_GT(frame_size_ms, 0);
RTC_DCHECK_EQ(frame_size_ms % 10, 0);
@@ -381,9 +388,14 @@
if (error == OPUS_OK && state->decoder) {
// Creation of memory all ok.
state->channels = channels;
- state->prev_decoded_samples = DefaultFrameSizePerChannel(sample_rate_hz);
- state->in_dtx_mode = 0;
state->sample_rate_hz = sample_rate_hz;
+ state->plc_use_prev_decoded_samples =
+ webrtc::field_trial::IsEnabled(kPlcUsePrevDecodedSamplesFieldTrial);
+ if (state->plc_use_prev_decoded_samples) {
+ state->prev_decoded_samples =
+ DefaultFrameSizePerChannel(state->sample_rate_hz);
+ }
+ state->in_dtx_mode = 0;
*inst = state;
return 0;
}
@@ -420,9 +432,14 @@
if (error == OPUS_OK && state->multistream_decoder) {
// Creation of memory all ok.
state->channels = channels;
- state->prev_decoded_samples = DefaultFrameSizePerChannel(48000);
- state->in_dtx_mode = 0;
state->sample_rate_hz = 48000;
+ state->plc_use_prev_decoded_samples =
+ webrtc::field_trial::IsEnabled(kPlcUsePrevDecodedSamplesFieldTrial);
+ if (state->plc_use_prev_decoded_samples) {
+ state->prev_decoded_samples =
+ DefaultFrameSizePerChannel(state->sample_rate_hz);
+ }
+ state->in_dtx_mode = 0;
*inst = state;
return 0;
}
@@ -517,17 +534,20 @@
static int DecodePlc(OpusDecInst* inst, int16_t* decoded) {
int16_t audio_type = 0;
int decoded_samples;
- int plc_samples;
+ int plc_samples =
+ FrameSizePerChannel(kWebRtcOpusPlcFrameSizeMs, inst->sample_rate_hz);
- /* The number of samples we ask for is |number_of_lost_frames| times
- * |prev_decoded_samples_|. Limit the number of samples to maximum
- * |MaxFrameSizePerChannel()|. */
- plc_samples = inst->prev_decoded_samples;
- const int max_samples_per_channel =
- MaxFrameSizePerChannel(inst->sample_rate_hz);
- plc_samples = plc_samples <= max_samples_per_channel
- ? plc_samples
- : max_samples_per_channel;
+ if (inst->plc_use_prev_decoded_samples) {
+ /* The number of samples we ask for is |number_of_lost_frames| times
+ * |prev_decoded_samples_|. Limit the number of samples to maximum
+ * |MaxFrameSizePerChannel()|. */
+ plc_samples = inst->prev_decoded_samples;
+ const int max_samples_per_channel =
+ MaxFrameSizePerChannel(inst->sample_rate_hz);
+ plc_samples = plc_samples <= max_samples_per_channel
+ ? plc_samples
+ : max_samples_per_channel;
+ }
decoded_samples =
DecodeNative(inst, NULL, 0, plc_samples, decoded, &audio_type, 0);
if (decoded_samples < 0) {
@@ -556,8 +576,10 @@
return -1;
}
- /* Update decoded sample memory, to be used by the PLC in case of losses. */
- inst->prev_decoded_samples = decoded_samples;
+ if (inst->plc_use_prev_decoded_samples) {
+ /* Update decoded sample memory, to be used by the PLC in case of losses. */
+ inst->prev_decoded_samples = decoded_samples;
+ }
return decoded_samples;
}
@@ -612,14 +634,17 @@
}
int WebRtcOpus_PlcDuration(OpusDecInst* inst) {
- /* The number of samples we ask for is |number_of_lost_frames| times
- * |prev_decoded_samples_|. Limit the number of samples to maximum
- * |MaxFrameSizePerChannel()|. */
- const int plc_samples = inst->prev_decoded_samples;
- const int max_samples_per_channel =
- MaxFrameSizePerChannel(inst->sample_rate_hz);
- return plc_samples <= max_samples_per_channel ? plc_samples
- : max_samples_per_channel;
+ if (inst->plc_use_prev_decoded_samples) {
+ /* The number of samples we ask for is |number_of_lost_frames| times
+ * |prev_decoded_samples_|. Limit the number of samples to maximum
+ * |MaxFrameSizePerChannel()|. */
+ const int plc_samples = inst->prev_decoded_samples;
+ const int max_samples_per_channel =
+ MaxFrameSizePerChannel(inst->sample_rate_hz);
+ return plc_samples <= max_samples_per_channel ? plc_samples
+ : max_samples_per_channel;
+ }
+ return FrameSizePerChannel(kWebRtcOpusPlcFrameSizeMs, inst->sample_rate_hz);
}
int WebRtcOpus_FecDurationEst(const uint8_t* payload,
diff --git a/modules/audio_coding/codecs/opus/opus_unittest.cc b/modules/audio_coding/codecs/opus/opus_unittest.cc
index 10897fb..0cc4f25 100644
--- a/modules/audio_coding/codecs/opus/opus_unittest.cc
+++ b/modules/audio_coding/codecs/opus/opus_unittest.cc
@@ -213,17 +213,34 @@
WebRtcOpusDecInst* decoder,
int16_t* output_audio,
int16_t* audio_type) {
+ const int input_samples_per_channel =
+ rtc::CheckedDivExact(input_audio.size(), channels_);
int encoded_bytes_int =
- WebRtcOpus_Encode(encoder, input_audio.data(),
- rtc::CheckedDivExact(input_audio.size(), channels_),
+ WebRtcOpus_Encode(encoder, input_audio.data(), input_samples_per_channel,
kMaxBytes, bitstream_);
EXPECT_GE(encoded_bytes_int, 0);
encoded_bytes_ = static_cast<size_t>(encoded_bytes_int);
- int est_len = WebRtcOpus_DurationEst(decoder, bitstream_, encoded_bytes_);
- int act_len = WebRtcOpus_Decode(decoder, bitstream_, encoded_bytes_,
- output_audio, audio_type);
- EXPECT_EQ(est_len, act_len);
- return act_len;
+ if (encoded_bytes_ != 0) {
+ int est_len = WebRtcOpus_DurationEst(decoder, bitstream_, encoded_bytes_);
+ int act_len = WebRtcOpus_Decode(decoder, bitstream_, encoded_bytes_,
+ output_audio, audio_type);
+ EXPECT_EQ(est_len, act_len);
+ return act_len;
+ } else {
+ int total_dtx_len = 0;
+ const int output_samples_per_channel = input_samples_per_channel *
+ decoder_sample_rate_hz_ /
+ encoder_sample_rate_hz_;
+ while (total_dtx_len < output_samples_per_channel) {
+ int est_len = WebRtcOpus_DurationEst(decoder, NULL, 0);
+ int act_len = WebRtcOpus_Decode(decoder, NULL, 0,
+ &output_audio[total_dtx_len * channels_],
+ audio_type);
+ EXPECT_EQ(est_len, act_len);
+ total_dtx_len += act_len;
+ }
+ return total_dtx_len;
+ }
}
// Test if encoder/decoder can enter DTX mode properly and do not enter DTX when
@@ -808,8 +825,10 @@
opus_decoder_, output_data_decode, &audio_type));
// Call decoder PLC.
- int16_t* plc_buffer = new int16_t[decode_samples_per_channel * channels_];
- EXPECT_EQ(decode_samples_per_channel,
+ constexpr int kPlcDurationMs = 10;
+ const int plc_samples = decoder_sample_rate_hz_ * kPlcDurationMs / 1000;
+ int16_t* plc_buffer = new int16_t[plc_samples * channels_];
+ EXPECT_EQ(plc_samples,
WebRtcOpus_Decode(opus_decoder_, NULL, 0, plc_buffer, &audio_type));
// Free memory.
diff --git a/modules/audio_coding/neteq/neteq_unittest.cc b/modules/audio_coding/neteq/neteq_unittest.cc
index a192611..58177dc 100644
--- a/modules/audio_coding/neteq/neteq_unittest.cc
+++ b/modules/audio_coding/neteq/neteq_unittest.cc
@@ -508,11 +508,11 @@
webrtc::test::ResourcePath("audio_coding/neteq_opus_dtx", "rtp");
const std::string maybe_sse =
- "713af6c92881f5aab1285765ee6680da9d1c06ce|"
- "2ac10c4e79aeedd0df2863b079da5848b40f00b5";
+ "0bdeb4ccf95a2577e38274360903ad099fc46787|"
+ "f7bbf5d92a0595a2a3445ffbaddfb20e98b6e94e";
const std::string output_checksum = PlatformChecksum(
- maybe_sse, "3ec991b96872123f1554c03c543ca5d518431e46",
- "da9f9a2d94e0c2d67342fad4965d7b91cda50b25", maybe_sse, maybe_sse);
+ maybe_sse, "6d200cc51a001b6137abf67db2bb8eeb0375cdee",
+ "36d43761de86b12520cf2e63f97372a2b7c6f939", maybe_sse, maybe_sse);
const std::string network_stats_checksum =
"8caf49765f35b6862066d3f17531ce44d8e25f60";
diff --git a/modules/audio_coding/test/opus_test.cc b/modules/audio_coding/test/opus_test.cc
index 10644e2..e110924 100644
--- a/modules/audio_coding/test/opus_test.cc
+++ b/modules/audio_coding/test/opus_test.cc
@@ -299,9 +299,19 @@
opus_mono_decoder_, bitstream, bitstream_len_byte,
&out_audio[decoded_samples * channels], &audio_type);
} else {
- decoded_samples += WebRtcOpus_Decode(
- opus_mono_decoder_, NULL, 0,
- &out_audio[decoded_samples * channels], &audio_type);
+ // Call decoder PLC.
+ constexpr int kPlcDurationMs = 10;
+ constexpr int kPlcSamples = 48 * kPlcDurationMs;
+ size_t total_plc_samples = 0;
+ while (total_plc_samples < frame_length) {
+ int ret = WebRtcOpus_Decode(
+ opus_mono_decoder_, NULL, 0,
+ &out_audio[decoded_samples * channels], &audio_type);
+ EXPECT_EQ(ret, kPlcSamples);
+ decoded_samples += ret;
+ total_plc_samples += ret;
+ }
+ EXPECT_EQ(total_plc_samples, frame_length);
}
} else {
if (!lost_packet) {
@@ -309,9 +319,19 @@
opus_stereo_decoder_, bitstream, bitstream_len_byte,
&out_audio[decoded_samples * channels], &audio_type);
} else {
- decoded_samples += WebRtcOpus_Decode(
- opus_stereo_decoder_, NULL, 0,
- &out_audio[decoded_samples * channels], &audio_type);
+ // Call decoder PLC.
+ constexpr int kPlcDurationMs = 10;
+ constexpr int kPlcSamples = 48 * kPlcDurationMs;
+ size_t total_plc_samples = 0;
+ while (total_plc_samples < frame_length) {
+ int ret = WebRtcOpus_Decode(
+ opus_stereo_decoder_, NULL, 0,
+ &out_audio[decoded_samples * channels], &audio_type);
+ EXPECT_EQ(ret, kPlcSamples);
+ decoded_samples += ret;
+ total_plc_samples += ret;
+ }
+ EXPECT_EQ(total_plc_samples, frame_length);
}
}
