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

Reland "Change buffer level filter to store current level in number of samples."

Browse Source 
This is a reland of 87977dd06e702ed517f26235c12e37bd927527c7

Original change's description:
> Change buffer level filter to store current level in number of samples.
> 
> The buffer level should not be converted back and forth between samples and packets in case of variable packet lengths.
> 
> Bug: webrtc:10736
> Change-Id: Ia08dcfac3d8104dc79fbad0704a5f6f12a050a01
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/142178
> Reviewed-by: Minyue Li <minyue@webrtc.org>
> Commit-Queue: Jakob Ivarsson <jakobi@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#28368}

Bug: webrtc:10736
Change-Id: I1ff603e65cdd31c7429f36b035dcc00a17b68f3b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/143787
Commit-Queue: Minyue Li <minyue@webrtc.org>
Reviewed-by: Minyue Li <minyue@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#28393}
This commit is contained in:
Jakob Ivarsson
2019-06-26 16:24:04 +02:00
committed by Commit Bot
parent 4d6951669c
commit 0ded32d5a3
9 changed files with 114 additions and 184 deletions
Download Patch File Download Diff File Expand all files Collapse all files

50
modules/audio_coding/acm2/audio_coding_module_unittest.cc
View File

@ -994,35 +994,35 @@ class AcmReceiverBitExactnessOldApi : public ::testing::Test {
#if (defined(WEBRTC_CODEC_ISAC) || defined(WEBRTC_CODEC_ISACFX)) && \ #if (defined(WEBRTC_CODEC_ISAC) || defined(WEBRTC_CODEC_ISACFX)) && \
defined(WEBRTC_CODEC_ILBC) defined(WEBRTC_CODEC_ILBC)
TEST_F(AcmReceiverBitExactnessOldApi, 8kHzOutput) { TEST_F(AcmReceiverBitExactnessOldApi, 8kHzOutput) {
Run(8000, PlatformChecksum("bcfbe2e89b4317b22e29557168edf187", Run(8000, PlatformChecksum("73e82368b90b0708bd970da1f357f71d",
"af15addb648cf7f032d6415672365fb3", "e777abcc66fccf8e86ac18450ad8b23c",
"54a0008eb79537dee1d8fdaa5bc29f4b", "5a668d4075a39cd07a2db82ec3bf19ba",
"4598140b5e4f7ee66c5adad609e65a3e", "4598140b5e4f7ee66c5adad609e65a3e",
"3155d7f2593a3276986f36221a61783c")); "99d17cc50d41232a4f96c976231cb59b"));
} }
TEST_F(AcmReceiverBitExactnessOldApi, 16kHzOutput) { TEST_F(AcmReceiverBitExactnessOldApi, 16kHzOutput) {
Run(16000, PlatformChecksum("1737deef193e6c90e139ce82b7361ae4", Run(16000, PlatformChecksum("f0b9d6961c243a3397b0bb95191b189b",
"9e2a9f7728c71d6559ce3a32d2b10a5d", "c73877b73a7ae2687eabc88de3d3f5bc",
"114958862099142ac78b12100c21cb8d", "70d24360be8290abbd0e56c38f83cdef",
"f2aad418af974a3b1694d5ae5cc2c3c7", "f2aad418af974a3b1694d5ae5cc2c3c7",
"af2889a5ca84fb40c9aa209b9318ee7a")); "564b1b5d2d9bcace5285623cd9822b57"));
} }
TEST_F(AcmReceiverBitExactnessOldApi, 32kHzOutput) { TEST_F(AcmReceiverBitExactnessOldApi, 32kHzOutput) {
Run(32000, PlatformChecksum("1bf40ff024c6aa5b832d1d242c29cb3b", Run(32000, PlatformChecksum("881a799ad91f845b1cd833e4e42d1791",
"3c9690cd136e9ecd1b26a22f70fe1d5c", "90e478af57f11bcf678b72ed1ba87765",
"a1a3a01d8e25fcd11f1cedcd02e968b8", "774657761e20fdec6d325d7d4b4101a7",
"100869c8dcde51346c2073e52a272d98", "100869c8dcde51346c2073e52a272d98",
"33695077e9ec6bca80819ce2ba263a78")); "4b77795ba2581097dc8e4db6e6a3a921"));
} }
TEST_F(AcmReceiverBitExactnessOldApi, 48kHzOutput) { TEST_F(AcmReceiverBitExactnessOldApi, 48kHzOutput) {
Run(48000, PlatformChecksum("bf92db1e502deff5adf6fd2e6ab9a2e5", Run(48000, PlatformChecksum("991b729aef7f08eca75d4c9ece848264",
"c37b110ab50d87620972daee5d1eaf31", "0334f53d4e96156edc302e46ff5cfaec",
"5d55b68be7bcf39b60fcc74519363fb4", "a578705020fe94ebde31b27d61035299",
"bd44bf97e7899186532f91235cef444d", "bd44bf97e7899186532f91235cef444d",
"32eec738698ffe62b9777d6a349cd596")); "c0d4185eacde6cd470c1a2ce4cd45318"));
} }
TEST_F(AcmReceiverBitExactnessOldApi, 48kHzOutputExternalDecoder) { TEST_F(AcmReceiverBitExactnessOldApi, 48kHzOutputExternalDecoder) {
@ -1105,11 +1105,11 @@ TEST_F(AcmReceiverBitExactnessOldApi, 48kHzOutputExternalDecoder) {
rtc::scoped_refptr<rtc::RefCountedObject<ADFactory>> factory( rtc::scoped_refptr<rtc::RefCountedObject<ADFactory>> factory(
new rtc::RefCountedObject<ADFactory>); new rtc::RefCountedObject<ADFactory>);
Run(48000, Run(48000,
PlatformChecksum("bf92db1e502deff5adf6fd2e6ab9a2e5", PlatformChecksum("991b729aef7f08eca75d4c9ece848264",
"c37b110ab50d87620972daee5d1eaf31", "0334f53d4e96156edc302e46ff5cfaec",
"5d55b68be7bcf39b60fcc74519363fb4", "a578705020fe94ebde31b27d61035299",
"bd44bf97e7899186532f91235cef444d", "bd44bf97e7899186532f91235cef444d",
"32eec738698ffe62b9777d6a349cd596"), "c0d4185eacde6cd470c1a2ce4cd45318"),
factory, [](AudioCodingModule* acm) { factory, [](AudioCodingModule* acm) {
acm->SetReceiveCodecs({{0, {"MockPCMu", 8000, 1}}, acm->SetReceiveCodecs({{0, {"MockPCMu", 8000, 1}},
{103, {"ISAC", 16000, 1}}, {103, {"ISAC", 16000, 1}},
@ -1328,7 +1328,7 @@ TEST_F(AcmSenderBitExactnessOldApi, IsacWb30ms) {
Run(AcmReceiverBitExactnessOldApi::PlatformChecksum( Run(AcmReceiverBitExactnessOldApi::PlatformChecksum(
"2c9cb15d4ed55b5a0cadd04883bc73b0", "2c9cb15d4ed55b5a0cadd04883bc73b0",
"9336a9b993cbd8a751f0e8958e66c89c", "9336a9b993cbd8a751f0e8958e66c89c",
"bd4682225f7c4ad5f2049f6769713ac2", "5c2eb46199994506236f68b2c8e51b0d",
"343f1f42be0607c61e6516aece424609", "343f1f42be0607c61e6516aece424609",
"2c9cb15d4ed55b5a0cadd04883bc73b0"), "2c9cb15d4ed55b5a0cadd04883bc73b0"),
AcmReceiverBitExactnessOldApi::PlatformChecksum( AcmReceiverBitExactnessOldApi::PlatformChecksum(
@ -1343,11 +1343,11 @@ TEST_F(AcmSenderBitExactnessOldApi, IsacWb30ms) {
TEST_F(AcmSenderBitExactnessOldApi, IsacWb60ms) { TEST_F(AcmSenderBitExactnessOldApi, IsacWb60ms) {
ASSERT_NO_FATAL_FAILURE(SetUpTest("ISAC", 16000, 1, 103, 960, 960)); ASSERT_NO_FATAL_FAILURE(SetUpTest("ISAC", 16000, 1, 103, 960, 960));
Run(AcmReceiverBitExactnessOldApi::PlatformChecksum( Run(AcmReceiverBitExactnessOldApi::PlatformChecksum(
"1ad29139a04782a33daad8c2b9b35875", "f59760fa000991ee5fa81f2e607db254",
"14d63c5f08127d280e722e3191b73bdd", "986aa16d7097a26e32e212e39ec58517",
"edcf26694c289e3d9691faf79b74f09f", "9a81e467eb1485f84aca796f8ea65011",
"ef75e900e6f375e3061163c53fd09a63", "ef75e900e6f375e3061163c53fd09a63",
"1ad29139a04782a33daad8c2b9b35875"), "f59760fa000991ee5fa81f2e607db254"),
AcmReceiverBitExactnessOldApi::PlatformChecksum( AcmReceiverBitExactnessOldApi::PlatformChecksum(
"9e0a0ab743ad987b55b8e14802769c56", "9e0a0ab743ad987b55b8e14802769c56",
"ebe04a819d3a9d83a83a17f271e1139a", "ebe04a819d3a9d83a83a17f271e1139a",

34
modules/audio_coding/neteq/buffer_level_filter.cc
View File

@ -26,32 +26,22 @@ void BufferLevelFilter::Reset() {
level_factor_ = 253; level_factor_ = 253;
} }
void BufferLevelFilter::Update(size_t buffer_size_packets, void BufferLevelFilter::Update(size_t buffer_size_samples,
int time_stretched_samples, int time_stretched_samples) {
size_t packet_len_samples) {
// Filter: // Filter:
// |filtered_current_level_| = |level_factor_| * |filtered_current_level_| + // |filtered_current_level_| = |level_factor_| * |filtered_current_level_| +
// (1 - |level_factor_|) * |buffer_size_packets| // (1 - |level_factor_|) * |buffer_size_samples|
// |level_factor_| and |filtered_current_level_| are in Q8. // |level_factor_| and |filtered_current_level_| are in Q8.
// |buffer_size_packets| is in Q0. // |buffer_size_samples| is in Q0.
filtered_current_level_ = filtered_current_level_ =
((level_factor_ * filtered_current_level_) >> 8) + ((level_factor_ * filtered_current_level_) >> 8) +
((256 - level_factor_) * rtc::dchecked_cast<int>(buffer_size_packets)); ((256 - level_factor_) * rtc::dchecked_cast<int>(buffer_size_samples));
// Account for time-scale operations (accelerate and pre-emptive expand). // Account for time-scale operations (accelerate and pre-emptive expand) and
if (time_stretched_samples && packet_len_samples > 0) { // make sure that the filtered value remains non-negative.
// Time-scaling has been performed since last filter update. Subtract the filtered_current_level_ = rtc::saturated_cast<int>(std::max<int64_t>(
// value of |time_stretched_samples| from |filtered_current_level_| after 0,
// converting |time_stretched_samples| from samples to packets in Q8. filtered_current_level_ - (int64_t{time_stretched_samples} * (1 << 8))));
// Make sure that the filtered value remains non-negative.
int64_t time_stretched_packets =
(int64_t{time_stretched_samples} * (1 << 8)) /
rtc::dchecked_cast<int64_t>(packet_len_samples);
filtered_current_level_ = rtc::saturated_cast<int>(
std::max<int64_t>(0, filtered_current_level_ - time_stretched_packets));
}
} }
void BufferLevelFilter::SetTargetBufferLevel(int target_buffer_level) { void BufferLevelFilter::SetTargetBufferLevel(int target_buffer_level) {
@ -66,8 +56,4 @@ void BufferLevelFilter::SetTargetBufferLevel(int target_buffer_level) {
} }
} }
int BufferLevelFilter::filtered_current_level() const {
return filtered_current_level_;
}
} // namespace webrtc } // namespace webrtc

20
modules/audio_coding/neteq/buffer_level_filter.h
View File

@ -24,20 +24,20 @@ class BufferLevelFilter {
virtual void Reset(); virtual void Reset();
// Updates the filter. Current buffer size is |buffer_size_packets| (Q0). // Updates the filter. Current buffer size is |buffer_size_packets| (Q0).
// If |time_stretched_samples| is non-zero, the value is converted to the // |time_stretched_samples| is subtracted from the filtered value (thus
// corresponding number of packets, and is subtracted from the filtered // bypassing the filter operation).
// value (thus bypassing the filter operation). |packet_len_samples| is the virtual void Update(size_t buffer_size_samples, int time_stretched_samples);
// number of audio samples carried in each incoming packet.
virtual void Update(size_t buffer_size_packets,
int time_stretched_samples,
size_t packet_len_samples);
// Set the current target buffer level (obtained from // Set the current target buffer level in number of packets (obtained from
// DelayManager::base_target_level()). Used to select the appropriate // DelayManager::base_target_level()). Used to select the appropriate
// filter coefficient. // filter coefficient.
virtual void SetTargetBufferLevel(int target_buffer_level); virtual void SetTargetBufferLevel(int target_buffer_level_packets);
virtual int filtered_current_level() const; // Returns filtered current level in number of samples.
virtual int filtered_current_level() const {
// Round to nearest whole sample.
return (filtered_current_level_ + (1 << 7)) >> 8;
}
private: private:
int level_factor_; // Filter factor for the buffer level filter in Q8. int level_factor_; // Filter factor for the buffer level filter in Q8.

85
modules/audio_coding/neteq/buffer_level_filter_unittest.cc
View File

@ -35,18 +35,17 @@ TEST(BufferLevelFilter, ConvergenceTest) {
ss << "times = " << times << ", value = " << value; ss << "times = " << times << ", value = " << value;
SCOPED_TRACE(ss.str()); // Print out the parameter values on failure. SCOPED_TRACE(ss.str()); // Print out the parameter values on failure.
for (int i = 0; i < times; ++i) { for (int i = 0; i < times; ++i) {
filter.Update(value, 0 /* time_stretched_samples */, filter.Update(value, 0 /* time_stretched_samples */);
160 /* packet_len_samples */);
} }
// Expect the filtered value to be (theoretically) // Expect the filtered value to be (theoretically)
// (1 - (251/256) ^ |times|) * |value|. // (1 - (251/256) ^ |times|) * |value|.
double expected_value_double = (1 - pow(251.0 / 256.0, times)) * value; double expected_value_double = (1 - pow(251.0 / 256.0, times)) * value;
int expected_value = static_cast<int>(expected_value_double); int expected_value = static_cast<int>(expected_value_double);
// filtered_current_level() returns the value in Q8.
// The actual value may differ slightly from the expected value due to // The actual value may differ slightly from the expected value due to
// intermediate-stage rounding errors in the filter implementation. // intermediate-stage rounding errors in the filter implementation.
// This is why we have to use EXPECT_NEAR with a tolerance of +/-1. // This is why we have to use EXPECT_NEAR with a tolerance of +/-1.
EXPECT_NEAR(expected_value, filter.filtered_current_level() >> 8, 1); EXPECT_NEAR(expected_value, filter.filtered_current_level(), 1);
} }
} }
} }
@ -60,38 +59,32 @@ TEST(BufferLevelFilter, FilterFactor) {
filter.SetTargetBufferLevel(3); // Makes filter coefficient 252/256. filter.SetTargetBufferLevel(3); // Makes filter coefficient 252/256.
for (int i = 0; i < kTimes; ++i) { for (int i = 0; i < kTimes; ++i) {
filter.Update(kValue, 0 /* time_stretched_samples */, filter.Update(kValue, 0 /* time_stretched_samples */);
160 /* packet_len_samples */);
} }
// Expect the filtered value to be // Expect the filtered value to be
// (1 - (252/256) ^ |kTimes|) * |kValue|. // (1 - (252/256) ^ |kTimes|) * |kValue|.
int expected_value = 14; int expected_value = 15;
// filtered_current_level() returns the value in Q8. EXPECT_EQ(expected_value, filter.filtered_current_level());
EXPECT_EQ(expected_value, filter.filtered_current_level() >> 8);
filter.Reset(); filter.Reset();
filter.SetTargetBufferLevel(7); // Makes filter coefficient 253/256. filter.SetTargetBufferLevel(7); // Makes filter coefficient 253/256.
for (int i = 0; i < kTimes; ++i) { for (int i = 0; i < kTimes; ++i) {
filter.Update(kValue, 0 /* time_stretched_samples */, filter.Update(kValue, 0 /* time_stretched_samples */);
160 /* packet_len_samples */);
} }
// Expect the filtered value to be // Expect the filtered value to be
// (1 - (253/256) ^ |kTimes|) * |kValue|. // (1 - (253/256) ^ |kTimes|) * |kValue|.
expected_value = 11; expected_value = 11;
// filtered_current_level() returns the value in Q8. EXPECT_EQ(expected_value, filter.filtered_current_level());
EXPECT_EQ(expected_value, filter.filtered_current_level() >> 8);
filter.Reset(); filter.Reset();
filter.SetTargetBufferLevel(8); // Makes filter coefficient 254/256. filter.SetTargetBufferLevel(8); // Makes filter coefficient 254/256.
for (int i = 0; i < kTimes; ++i) { for (int i = 0; i < kTimes; ++i) {
filter.Update(kValue, 0 /* time_stretched_samples */, filter.Update(kValue, 0 /* time_stretched_samples */);
160 /* packet_len_samples */);
} }
// Expect the filtered value to be // Expect the filtered value to be
// (1 - (254/256) ^ |kTimes|) * |kValue|. // (1 - (254/256) ^ |kTimes|) * |kValue|.
expected_value = 7; expected_value = 8;
// filtered_current_level() returns the value in Q8. EXPECT_EQ(expected_value, filter.filtered_current_level());
EXPECT_EQ(expected_value, filter.filtered_current_level() >> 8);
} }
TEST(BufferLevelFilter, TimeStretchedSamples) { TEST(BufferLevelFilter, TimeStretchedSamples) {
@ -100,62 +93,24 @@ TEST(BufferLevelFilter, TimeStretchedSamples) {
// Update 10 times with value 100. // Update 10 times with value 100.
const int kTimes = 10; const int kTimes = 10;
const int kValue = 100; const int kValue = 100;
const int kPacketSizeSamples = 160; const int kTimeStretchedSamples = 3;
const int kNumPacketsStretched = 2;
const int kTimeStretchedSamples = kNumPacketsStretched * kPacketSizeSamples;
for (int i = 0; i < kTimes; ++i) { for (int i = 0; i < kTimes; ++i) {
// Packet size set to 0. Do not expect the parameter filter.Update(kValue, 0);
// |kTimeStretchedSamples| to have any effect.
filter.Update(kValue, kTimeStretchedSamples, 0 /* packet_len_samples */);
} }
// Expect the filtered value to be // Expect the filtered value to be
// (1 - (251/256) ^ |kTimes|) * |kValue|. // (1 - (251/256) ^ |kTimes|) * |kValue|.
const int kExpectedValue = 17; const int kExpectedValue = 18;
// filtered_current_level() returns the value in Q8. EXPECT_EQ(kExpectedValue, filter.filtered_current_level());
EXPECT_EQ(kExpectedValue, filter.filtered_current_level() >> 8);
// Update filter again, now with non-zero value for packet length. // Update filter again, now with non-zero value for packet length.
// Set the current filtered value to be the input, in order to isolate the // Set the current filtered value to be the input, in order to isolate the
// impact of |kTimeStretchedSamples|. // impact of |kTimeStretchedSamples|.
filter.Update(filter.filtered_current_level() >> 8, kTimeStretchedSamples, filter.Update(filter.filtered_current_level(), kTimeStretchedSamples);
kPacketSizeSamples); EXPECT_EQ(kExpectedValue - kTimeStretchedSamples,
EXPECT_EQ(kExpectedValue - kNumPacketsStretched, filter.filtered_current_level());
filter.filtered_current_level() >> 8);
// Try negative value and verify that we come back to the previous result. // Try negative value and verify that we come back to the previous result.
filter.Update(filter.filtered_current_level() >> 8, -kTimeStretchedSamples, filter.Update(filter.filtered_current_level(), -kTimeStretchedSamples);
kPacketSizeSamples); EXPECT_EQ(kExpectedValue, filter.filtered_current_level());
EXPECT_EQ(kExpectedValue, filter.filtered_current_level() >> 8);
}
TEST(BufferLevelFilter, TimeStretchedSamplesNegativeUnevenFrames) {
BufferLevelFilter filter;
filter.SetTargetBufferLevel(1); // Makes filter coefficient 251/256.
// Update 10 times with value 100.
const int kTimes = 10;
const int kValue = 100;
const int kPacketSizeSamples = 160;
const int kTimeStretchedSamples = -3.1415 * kPacketSizeSamples;
for (int i = 0; i < kTimes; ++i) {
// Packet size set to 0. Do not expect the parameter
// |kTimeStretchedSamples| to have any effect.
filter.Update(kValue, kTimeStretchedSamples, 0 /* packet_len_samples */);
}
// Expect the filtered value to be
// (1 - (251/256) ^ |kTimes|) * |kValue|.
const int kExpectedValue = 17;
// filtered_current_level() returns the value in Q8.
EXPECT_EQ(kExpectedValue, filter.filtered_current_level() >> 8);
// Update filter again, now with non-zero value for packet length.
// Set the current filtered value to be the input, in order to isolate the
// impact of |kTimeStretchedSamples|.
filter.Update(filter.filtered_current_level() >> 8, kTimeStretchedSamples,
kPacketSizeSamples);
EXPECT_EQ(21, filter.filtered_current_level() >> 8);
// Try negative value and verify that we come back to the previous result.
filter.Update(filter.filtered_current_level() >> 8, -kTimeStretchedSamples,
kPacketSizeSamples);
EXPECT_EQ(kExpectedValue, filter.filtered_current_level() >> 8);
} }
} // namespace webrtc } // namespace webrtc

51
modules/audio_coding/neteq/decision_logic.cc
View File

@ -113,11 +113,9 @@ Operations DecisionLogic::GetDecision(const SyncBuffer& sync_buffer,
cng_state_ = kCngInternalOn; cng_state_ = kCngInternalOn;
} }
const size_t samples_left =
sync_buffer.FutureLength() - expand.overlap_length();
// TODO(jakobi): Use buffer span instead of num samples. // TODO(jakobi): Use buffer span instead of num samples.
const size_t cur_size_samples = const size_t cur_size_samples =
samples_left + packet_buffer_.NumSamplesInBuffer(decoder_frame_length); packet_buffer_.NumSamplesInBuffer(decoder_frame_length);
prev_time_scale_ = prev_time_scale_ =
prev_time_scale_ && (prev_mode == kModeAccelerateSuccess || prev_time_scale_ && (prev_mode == kModeAccelerateSuccess ||
@ -175,8 +173,7 @@ Operations DecisionLogic::GetDecision(const SyncBuffer& sync_buffer,
// if the mute factor is low enough (otherwise the expansion was short enough // if the mute factor is low enough (otherwise the expansion was short enough
// to not be noticable). // to not be noticable).
// Note that the MuteFactor is in Q14, so a value of 16384 corresponds to 1. // Note that the MuteFactor is in Q14, so a value of 16384 corresponds to 1.
size_t current_span = size_t current_span = packet_buffer_.GetSpanSamples(decoder_frame_length);
samples_left + packet_buffer_.GetSpanSamples(decoder_frame_length);
if ((prev_mode == kModeExpand || prev_mode == kModeCodecPlc) && if ((prev_mode == kModeExpand || prev_mode == kModeCodecPlc) &&
expand.MuteFactor(0) < 16384 / 2 && expand.MuteFactor(0) < 16384 / 2 &&
current_span < static_cast<size_t>(delay_manager_->TargetLevel() * current_span < static_cast<size_t>(delay_manager_->TargetLevel() *
@ -193,9 +190,9 @@ Operations DecisionLogic::GetDecision(const SyncBuffer& sync_buffer,
return ExpectedPacketAvailable(prev_mode, play_dtmf); return ExpectedPacketAvailable(prev_mode, play_dtmf);
} else if (!PacketBuffer::IsObsoleteTimestamp( } else if (!PacketBuffer::IsObsoleteTimestamp(
available_timestamp, target_timestamp, five_seconds_samples)) { available_timestamp, target_timestamp, five_seconds_samples)) {
return FuturePacketAvailable( return FuturePacketAvailable(decoder_frame_length, prev_mode,
sync_buffer, expand, decoder_frame_length, prev_mode, target_timestamp, target_timestamp, available_timestamp,
available_timestamp, play_dtmf, generated_noise_samples); play_dtmf, generated_noise_samples);
} else { } else {
// This implies that available_timestamp < target_timestamp, which can // This implies that available_timestamp < target_timestamp, which can
// happen when a new stream or codec is received. Signal for a reset. // happen when a new stream or codec is received. Signal for a reset.
@ -215,19 +212,13 @@ void DecisionLogic::FilterBufferLevel(size_t buffer_size_samples) {
buffer_level_filter_->SetTargetBufferLevel( buffer_level_filter_->SetTargetBufferLevel(
delay_manager_->base_target_level()); delay_manager_->base_target_level());
size_t buffer_size_packets = 0;
if (packet_length_samples_ > 0) {
// Calculate size in packets.
buffer_size_packets = buffer_size_samples / packet_length_samples_;
}
int sample_memory_local = 0; int sample_memory_local = 0;
if (prev_time_scale_) { if (prev_time_scale_) {
sample_memory_local = sample_memory_; sample_memory_local = sample_memory_;
timescale_countdown_ = tick_timer_->GetNewCountdown(kMinTimescaleInterval); timescale_countdown_ = tick_timer_->GetNewCountdown(kMinTimescaleInterval);
} }
buffer_level_filter_->Update(buffer_size_packets, sample_memory_local, buffer_level_filter_->Update(buffer_size_samples, sample_memory_local);
packet_length_samples_);
prev_time_scale_ = false; prev_time_scale_ = false;
} }
@ -283,15 +274,22 @@ Operations DecisionLogic::NoPacket(bool play_dtmf) {
Operations DecisionLogic::ExpectedPacketAvailable(Modes prev_mode, Operations DecisionLogic::ExpectedPacketAvailable(Modes prev_mode,
bool play_dtmf) { bool play_dtmf) {
if (!disallow_time_stretching_ && prev_mode != kModeExpand && !play_dtmf) { if (!disallow_time_stretching_ && prev_mode != kModeExpand && !play_dtmf) {
// Check criterion for time-stretching. // Check criterion for time-stretching. The values are in number of packets
// in Q8.
int low_limit, high_limit; int low_limit, high_limit;
delay_manager_->BufferLimits(&low_limit, &high_limit); delay_manager_->BufferLimits(&low_limit, &high_limit);
if (buffer_level_filter_->filtered_current_level() >= high_limit << 2) int buffer_level_packets = 0;
if (packet_length_samples_ > 0) {
buffer_level_packets =
((1 << 8) * buffer_level_filter_->filtered_current_level()) /
packet_length_samples_;
}
if (buffer_level_packets >= high_limit << 2)
return kFastAccelerate; return kFastAccelerate;
if (TimescaleAllowed()) { if (TimescaleAllowed()) {
if (buffer_level_filter_->filtered_current_level() >= high_limit) if (buffer_level_packets >= high_limit)
return kAccelerate; return kAccelerate;
if (buffer_level_filter_->filtered_current_level() < low_limit) if (buffer_level_packets < low_limit)
return kPreemptiveExpand; return kPreemptiveExpand;
} }
} }
@ -299,8 +297,6 @@ Operations DecisionLogic::ExpectedPacketAvailable(Modes prev_mode,
} }
Operations DecisionLogic::FuturePacketAvailable( Operations DecisionLogic::FuturePacketAvailable(
const SyncBuffer& sync_buffer,
const Expand& expand,
size_t decoder_frame_length, size_t decoder_frame_length,
Modes prev_mode, Modes prev_mode,
uint32_t target_timestamp, uint32_t target_timestamp,
@ -327,10 +323,8 @@ Operations DecisionLogic::FuturePacketAvailable(
return kNormal; return kNormal;
} }
const size_t samples_left =
sync_buffer.FutureLength() - expand.overlap_length();
const size_t cur_size_samples = const size_t cur_size_samples =
samples_left + packet_buffer_.NumPacketsInBuffer() * decoder_frame_length; packet_buffer_.NumPacketsInBuffer() * decoder_frame_length;
// If previous was comfort noise, then no merge is needed. // If previous was comfort noise, then no merge is needed.
if (prev_mode == kModeRfc3389Cng || prev_mode == kModeCodecInternalCng) { if (prev_mode == kModeRfc3389Cng || prev_mode == kModeCodecInternalCng) {
@ -365,8 +359,13 @@ Operations DecisionLogic::FuturePacketAvailable(
} }
bool DecisionLogic::UnderTargetLevel() const { bool DecisionLogic::UnderTargetLevel() const {
return buffer_level_filter_->filtered_current_level() <= int buffer_level_packets = 0;
delay_manager_->TargetLevel(); if (packet_length_samples_ > 0) {
buffer_level_packets =
((1 << 8) * buffer_level_filter_->filtered_current_level()) /
packet_length_samples_;
}
return buffer_level_packets <= delay_manager_->TargetLevel();
} }
bool DecisionLogic::ReinitAfterExpands(uint32_t timestamp_leap) const { bool DecisionLogic::ReinitAfterExpands(uint32_t timestamp_leap) const {

4
modules/audio_coding/neteq/decision_logic.h
View File

@ -134,9 +134,7 @@ class DecisionLogic final {
// Returns the operation to do given that the expected packet is not // Returns the operation to do given that the expected packet is not
// available, but a packet further into the future is at hand. // available, but a packet further into the future is at hand.
Operations FuturePacketAvailable(const SyncBuffer& sync_buffer, Operations FuturePacketAvailable(size_t decoder_frame_length,
const Expand& expand,
size_t decoder_frame_length,
Modes prev_mode, Modes prev_mode,
uint32_t target_timestamp, uint32_t target_timestamp,
uint32_t available_timestamp, uint32_t available_timestamp,

6
modules/audio_coding/neteq/mock/mock_buffer_level_filter.h
View File

@ -22,10 +22,8 @@ class MockBufferLevelFilter : public BufferLevelFilter {
virtual ~MockBufferLevelFilter() { Die(); } virtual ~MockBufferLevelFilter() { Die(); }
MOCK_METHOD0(Die, void()); MOCK_METHOD0(Die, void());
MOCK_METHOD0(Reset, void()); MOCK_METHOD0(Reset, void());
MOCK_METHOD3(Update, MOCK_METHOD2(Update,
void(size_t buffer_size_packets, void(size_t buffer_size_samples, int time_stretched_samples));
int time_stretched_samples,
size_t packet_len_samples));
MOCK_METHOD1(SetTargetBufferLevel, void(int target_buffer_level)); MOCK_METHOD1(SetTargetBufferLevel, void(int target_buffer_level));
MOCK_CONST_METHOD0(filtered_current_level, int()); MOCK_CONST_METHOD0(filtered_current_level, int());
}; };

14
modules/audio_coding/neteq/neteq_impl.cc
View File

@ -310,18 +310,12 @@ int NetEqImpl::TargetDelayMs() const {
int NetEqImpl::FilteredCurrentDelayMs() const { int NetEqImpl::FilteredCurrentDelayMs() const {
rtc::CritScope lock(&crit_sect_); rtc::CritScope lock(&crit_sect_);
// Calculate the filtered packet buffer level in samples. The value from
// |buffer_level_filter_| is in number of packets, represented in Q8.
const size_t packet_buffer_samples =
(buffer_level_filter_->filtered_current_level() *
decoder_frame_length_) >>
8;
// Sum up the filtered packet buffer level with the future length of the sync // Sum up the filtered packet buffer level with the future length of the sync
// buffer, and divide the sum by the sample rate. // buffer.
const size_t delay_samples = const int delay_samples = buffer_level_filter_->filtered_current_level() +
packet_buffer_samples + sync_buffer_->FutureLength(); sync_buffer_->FutureLength();
// The division below will truncate. The return value is in ms. // The division below will truncate. The return value is in ms.
return static_cast<int>(delay_samples) / rtc::CheckedDivExact(fs_hz_, 1000); return delay_samples / rtc::CheckedDivExact(fs_hz_, 1000);
} }
int NetEqImpl::NetworkStatistics(NetEqNetworkStatistics* stats) { int NetEqImpl::NetworkStatistics(NetEqNetworkStatistics* stats) {

34
modules/audio_coding/neteq/neteq_unittest.cc
View File

@ -458,16 +458,16 @@ TEST_F(NetEqDecodingTest, MAYBE_TestBitExactness) {
webrtc::test::ResourcePath("audio_coding/neteq_universal_new", "rtp"); webrtc::test::ResourcePath("audio_coding/neteq_universal_new", "rtp");
const std::string output_checksum = const std::string output_checksum =
PlatformChecksum("9652cee1d6771a9cbfda821ae1bbdb41b0dd4dee", PlatformChecksum("998be2e5a707e636af0b6298f54bedfabe72aae1",
"54a7e32f163663c0af35bf70bf45cefc24ad62ef", "not used", "61e238ece4cd3b67d66a0b7047e06b20607dcb79", "not used",
"9652cee1d6771a9cbfda821ae1bbdb41b0dd4dee", "998be2e5a707e636af0b6298f54bedfabe72aae1",
"79496b0a1ef0a3824f3ee04789748a461bed643f"); "4116ac2a6e75baac3194b712d6fabe28b384275e");
const std::string network_stats_checksum = const std::string network_stats_checksum =
PlatformChecksum("c59b1f9f282b6d8733cdff975e3c150ca4a47d51", PlatformChecksum("3689c9f0ab9e50cefab3e44c37c3d7aa0de82ca4",
"bca95e565996a4ffd6e2ac15736e08843bdca93b", "not used", "0a596217fccd8d90eff7d1666b8cc63143eeda12", "not used",
"c59b1f9f282b6d8733cdff975e3c150ca4a47d51", "3689c9f0ab9e50cefab3e44c37c3d7aa0de82ca4",
"c59b1f9f282b6d8733cdff975e3c150ca4a47d51"); "3689c9f0ab9e50cefab3e44c37c3d7aa0de82ca4");
DecodeAndCompare(input_rtp_file, output_checksum, network_stats_checksum, DecodeAndCompare(input_rtp_file, output_checksum, network_stats_checksum,
FLAG_gen_ref); FLAG_gen_ref);
@ -486,17 +486,17 @@ TEST_F(NetEqDecodingTest, MAYBE_TestOpusBitExactness) {
// Checksum depends on libopus being compiled with or without SSE. // Checksum depends on libopus being compiled with or without SSE.
const std::string maybe_sse = const std::string maybe_sse =
"14a63b3c7b925c82296be4bafc71bec85f2915c2|" "14a63b3c7b925c82296be4bafc71bec85f2915c2|"
"2c05677daa968d6c68b92adf4affb7cd9bb4d363"; "eb0b68bddcac00fc85403df64f83126f8ea9bc93";
const std::string output_checksum = PlatformChecksum( const std::string output_checksum = PlatformChecksum(
maybe_sse, "b7b7ed802b0e18ee416973bf3b9ae98599b0181d", maybe_sse, "f95f2a220c9ca5d60b81c4653d46e0de2bee159f",
"5876e52dda90d5ca433c3726555b907b97c86374", maybe_sse, maybe_sse); "6f288a03d34958f62496f18fa85655593eef4dbe", maybe_sse, maybe_sse);
const std::string network_stats_checksum = const std::string network_stats_checksum =
PlatformChecksum("adb3272498e436d1c019cbfd71610e9510c54497", PlatformChecksum("0b3d34baffaf651812ffaf06ea1b5ce45ea1c47a",
"fa935a91abc7291db47428a2d7c5361b98713a92", "a71dce66c7bea85ba22d4e29a5298f606f810444",
"42106aa5267300f709f63737707ef07afd9dac61", "7c64e1e915bace7c4bf583484efd64eaf234552f",
"adb3272498e436d1c019cbfd71610e9510c54497", "0b3d34baffaf651812ffaf06ea1b5ce45ea1c47a",
"adb3272498e436d1c019cbfd71610e9510c54497"); "0b3d34baffaf651812ffaf06ea1b5ce45ea1c47a");
DecodeAndCompare(input_rtp_file, output_checksum, network_stats_checksum, DecodeAndCompare(input_rtp_file, output_checksum, network_stats_checksum,
FLAG_gen_ref); FLAG_gen_ref);
@ -796,7 +796,7 @@ TEST_F(NetEqDecodingTest, LongCngWithNegativeClockDriftNetworkFreeze) {
const double kDriftFactor = 1000.0 / (1000.0 + 25.0); const double kDriftFactor = 1000.0 / (1000.0 + 25.0);
const double kNetworkFreezeTimeMs = 5000.0; const double kNetworkFreezeTimeMs = 5000.0;
const bool kGetAudioDuringFreezeRecovery = false; const bool kGetAudioDuringFreezeRecovery = false;
const int kDelayToleranceMs = 50; const int kDelayToleranceMs = 60;
const int kMaxTimeToSpeechMs = 200; const int kMaxTimeToSpeechMs = 200;
LongCngWithClockDrift(kDriftFactor, kNetworkFreezeTimeMs, LongCngWithClockDrift(kDriftFactor, kNetworkFreezeTimeMs,
kGetAudioDuringFreezeRecovery, kDelayToleranceMs, kGetAudioDuringFreezeRecovery, kDelayToleranceMs,