
They make it possible to send bandwidth estimation info from decoder to encoder even if they are separate objects (which we want them to be because multithreading). R=henrik.lundin@webrtc.org Review URL: https://codereview.webrtc.org/1208923002. Cr-Commit-Position: refs/heads/master@{#9535}
272 lines
9.4 KiB
C++
272 lines
9.4 KiB
C++
/*
|
|
* Copyright (c) 2015 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 <algorithm>
|
|
#include <numeric>
|
|
#include <sstream>
|
|
#include <vector>
|
|
|
|
#include "testing/gtest/include/gtest/gtest.h"
|
|
#include "webrtc/base/buffer.h"
|
|
#include "webrtc/modules/audio_coding/codecs/isac/fix/interface/audio_encoder_isacfix.h"
|
|
#include "webrtc/modules/audio_coding/codecs/isac/main/interface/audio_encoder_isac.h"
|
|
#include "webrtc/modules/audio_coding/neteq/tools/input_audio_file.h"
|
|
#include "webrtc/test/testsupport/fileutils.h"
|
|
|
|
namespace webrtc {
|
|
|
|
namespace {
|
|
|
|
std::vector<int16_t> LoadSpeechData() {
|
|
webrtc::test::InputAudioFile input_file(
|
|
webrtc::test::ResourcePath("audio_coding/testfile32kHz", "pcm"));
|
|
static const int kIsacNumberOfSamples = 32 * 60; // 60 ms at 32 kHz
|
|
std::vector<int16_t> speech_data(kIsacNumberOfSamples);
|
|
input_file.Read(kIsacNumberOfSamples, speech_data.data());
|
|
return speech_data;
|
|
}
|
|
|
|
template <typename T>
|
|
IsacBandwidthInfo GetBwInfo(typename T::instance_type* inst) {
|
|
IsacBandwidthInfo bi;
|
|
T::GetBandwidthInfo(inst, &bi);
|
|
EXPECT_TRUE(bi.in_use);
|
|
return bi;
|
|
}
|
|
|
|
template <typename T>
|
|
rtc::Buffer EncodePacket(typename T::instance_type* inst,
|
|
const IsacBandwidthInfo* bi,
|
|
const int16_t* speech_data,
|
|
int framesize_ms) {
|
|
rtc::Buffer output(1000);
|
|
for (int i = 0;; ++i) {
|
|
if (bi)
|
|
T::SetBandwidthInfo(inst, bi);
|
|
int encoded_bytes = T::Encode(inst, speech_data, output.data());
|
|
if (i + 1 == framesize_ms / 10) {
|
|
EXPECT_GT(encoded_bytes, 0);
|
|
EXPECT_LE(static_cast<size_t>(encoded_bytes), output.size());
|
|
output.SetSize(encoded_bytes);
|
|
return output;
|
|
}
|
|
EXPECT_EQ(0, encoded_bytes);
|
|
}
|
|
}
|
|
|
|
class BoundedCapacityChannel final {
|
|
public:
|
|
BoundedCapacityChannel(int rate_bits_per_second)
|
|
: current_time_rtp_(0),
|
|
channel_rate_bytes_per_sample_(rate_bits_per_second /
|
|
(8.0 * kSamplesPerSecond)) {}
|
|
|
|
// Simulate sending the given number of bytes at the given RTP time. Returns
|
|
// the new current RTP time after the sending is done.
|
|
int Send(int send_time_rtp, int nbytes) {
|
|
current_time_rtp_ = std::max(current_time_rtp_, send_time_rtp) +
|
|
nbytes / channel_rate_bytes_per_sample_;
|
|
return current_time_rtp_;
|
|
}
|
|
|
|
private:
|
|
int current_time_rtp_;
|
|
// The somewhat strange unit for channel rate, bytes per sample, is because
|
|
// RTP time is measured in samples:
|
|
const double channel_rate_bytes_per_sample_;
|
|
static const int kSamplesPerSecond = 16000;
|
|
};
|
|
|
|
template <typename T, bool adaptive>
|
|
struct TestParam {};
|
|
|
|
template <>
|
|
struct TestParam<IsacFloat, true> {
|
|
static const int time_to_settle = 200;
|
|
static int ExpectedRateBitsPerSecond(int rate_bits_per_second) {
|
|
return rate_bits_per_second;
|
|
}
|
|
};
|
|
|
|
template <>
|
|
struct TestParam<IsacFix, true> {
|
|
static const int time_to_settle = 350;
|
|
static int ExpectedRateBitsPerSecond(int rate_bits_per_second) {
|
|
// For some reason, IsacFix fails to adapt to the channel's actual
|
|
// bandwidth. Instead, it settles on a few hundred packets at 10kbit/s,
|
|
// then a few hundred at 5kbit/s, then a few hundred at 10kbit/s, and so
|
|
// on. The 200 packets starting at 350 are in the middle of the first
|
|
// 10kbit/s run.
|
|
return 10000;
|
|
}
|
|
};
|
|
|
|
template <>
|
|
struct TestParam<IsacFloat, false> {
|
|
static const int time_to_settle = 0;
|
|
static int ExpectedRateBitsPerSecond(int rate_bits_per_second) {
|
|
return 32000;
|
|
}
|
|
};
|
|
|
|
template <>
|
|
struct TestParam<IsacFix, false> {
|
|
static const int time_to_settle = 0;
|
|
static int ExpectedRateBitsPerSecond(int rate_bits_per_second) {
|
|
return 16000;
|
|
}
|
|
};
|
|
|
|
// Test that the iSAC encoder produces identical output whether or not we use a
|
|
// conjoined encoder+decoder pair or a separate encoder and decoder that
|
|
// communicate BW estimation info explicitly.
|
|
template <typename T, bool adaptive>
|
|
void TestGetSetBandwidthInfo(const int16_t* speech_data,
|
|
int rate_bits_per_second) {
|
|
using Param = TestParam<T, adaptive>;
|
|
const int framesize_ms = adaptive ? 60 : 30;
|
|
|
|
// Conjoined encoder/decoder pair:
|
|
typename T::instance_type* encdec;
|
|
ASSERT_EQ(0, T::Create(&encdec));
|
|
ASSERT_EQ(0, T::EncoderInit(encdec, adaptive ? 0 : 1));
|
|
ASSERT_EQ(0, T::DecoderInit(encdec));
|
|
|
|
// Disjoint encoder/decoder pair:
|
|
typename T::instance_type* enc;
|
|
ASSERT_EQ(0, T::Create(&enc));
|
|
ASSERT_EQ(0, T::EncoderInit(enc, adaptive ? 0 : 1));
|
|
typename T::instance_type* dec;
|
|
ASSERT_EQ(0, T::Create(&dec));
|
|
ASSERT_EQ(0, T::DecoderInit(dec));
|
|
|
|
// 0. Get initial BW info from decoder.
|
|
auto bi = GetBwInfo<T>(dec);
|
|
|
|
BoundedCapacityChannel channel1(rate_bits_per_second),
|
|
channel2(rate_bits_per_second);
|
|
std::vector<size_t> packet_sizes;
|
|
for (int i = 0; i < Param::time_to_settle + 200; ++i) {
|
|
std::ostringstream ss;
|
|
ss << " i = " << i;
|
|
SCOPED_TRACE(ss.str());
|
|
|
|
// 1. Encode 6 * 10 ms (adaptive) or 3 * 10 ms (nonadaptive). The separate
|
|
// encoder is given the BW info before each encode call.
|
|
auto bitstream1 =
|
|
EncodePacket<T>(encdec, nullptr, speech_data, framesize_ms);
|
|
auto bitstream2 = EncodePacket<T>(enc, &bi, speech_data, framesize_ms);
|
|
EXPECT_EQ(bitstream1, bitstream2);
|
|
if (i > Param::time_to_settle)
|
|
packet_sizes.push_back(bitstream1.size());
|
|
|
|
// 2. Deliver the encoded data to the decoders (but don't actually ask them
|
|
// to decode it; that's not necessary). Then get new BW info from the
|
|
// separate decoder.
|
|
const int samples_per_packet = 16 * framesize_ms;
|
|
const int send_time = i * samples_per_packet;
|
|
EXPECT_EQ(0, T::UpdateBwEstimate(
|
|
encdec, bitstream1.data(), bitstream1.size(), i, send_time,
|
|
channel1.Send(send_time, bitstream1.size())));
|
|
EXPECT_EQ(0, T::UpdateBwEstimate(
|
|
dec, bitstream2.data(), bitstream2.size(), i, send_time,
|
|
channel2.Send(send_time, bitstream2.size())));
|
|
bi = GetBwInfo<T>(dec);
|
|
}
|
|
|
|
EXPECT_EQ(0, T::Free(encdec));
|
|
EXPECT_EQ(0, T::Free(enc));
|
|
EXPECT_EQ(0, T::Free(dec));
|
|
|
|
// The average send bitrate is close to the channel's capacity.
|
|
double avg_size =
|
|
std::accumulate(packet_sizes.begin(), packet_sizes.end(), 0) /
|
|
static_cast<double>(packet_sizes.size());
|
|
double avg_rate_bits_per_second = 8.0 * avg_size / (framesize_ms * 1e-3);
|
|
double expected_rate_bits_per_second =
|
|
Param::ExpectedRateBitsPerSecond(rate_bits_per_second);
|
|
EXPECT_GT(avg_rate_bits_per_second / expected_rate_bits_per_second, 0.95);
|
|
EXPECT_LT(avg_rate_bits_per_second / expected_rate_bits_per_second, 1.06);
|
|
|
|
// The largest packet isn't that large, and the smallest not that small.
|
|
size_t min_size = *std::min_element(packet_sizes.begin(), packet_sizes.end());
|
|
size_t max_size = *std::max_element(packet_sizes.begin(), packet_sizes.end());
|
|
double size_range = max_size - min_size;
|
|
EXPECT_LE(size_range / avg_size, 0.16);
|
|
}
|
|
|
|
} // namespace
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat12kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFloat, true>(LoadSpeechData().data(), 12000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat15kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFloat, true>(LoadSpeechData().data(), 15000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat19kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFloat, true>(LoadSpeechData().data(), 19000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat22kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFloat, true>(LoadSpeechData().data(), 22000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix12kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFix, true>(LoadSpeechData().data(), 12000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix15kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFix, true>(LoadSpeechData().data(), 15000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix19kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFix, true>(LoadSpeechData().data(), 19000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix22kAdaptive) {
|
|
TestGetSetBandwidthInfo<IsacFix, true>(LoadSpeechData().data(), 22000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat12k) {
|
|
TestGetSetBandwidthInfo<IsacFloat, false>(LoadSpeechData().data(), 12000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat15k) {
|
|
TestGetSetBandwidthInfo<IsacFloat, false>(LoadSpeechData().data(), 15000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat19k) {
|
|
TestGetSetBandwidthInfo<IsacFloat, false>(LoadSpeechData().data(), 19000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFloat22k) {
|
|
TestGetSetBandwidthInfo<IsacFloat, false>(LoadSpeechData().data(), 22000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix12k) {
|
|
TestGetSetBandwidthInfo<IsacFix, false>(LoadSpeechData().data(), 12000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix15k) {
|
|
TestGetSetBandwidthInfo<IsacFix, false>(LoadSpeechData().data(), 15000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix19k) {
|
|
TestGetSetBandwidthInfo<IsacFix, false>(LoadSpeechData().data(), 19000);
|
|
}
|
|
|
|
TEST(IsacCommonTest, GetSetBandwidthInfoFix22k) {
|
|
TestGetSetBandwidthInfo<IsacFix, false>(LoadSpeechData().data(), 22000);
|
|
}
|
|
|
|
} // namespace webrtc
|