Index: webrtc/modules/audio_coding/codecs/opus/opus_unittest.cc |
diff --git a/webrtc/modules/audio_coding/codecs/opus/opus_unittest.cc b/webrtc/modules/audio_coding/codecs/opus/opus_unittest.cc |
index c059fc5d0171f5952dbaaf1a44b4a895f160329e..fc5d84190b8aaff891399d73ff4b5cedde7c8080 100644 |
--- a/webrtc/modules/audio_coding/codecs/opus/opus_unittest.cc |
+++ b/webrtc/modules/audio_coding/codecs/opus/opus_unittest.cc |
@@ -36,7 +36,7 @@ class OpusTest : public TestWithParam<::testing::tuple<int, int>> { |
protected: |
OpusTest(); |
- void TestDtxEffect(bool dtx); |
+ void TestDtxEffect(bool dtx, int block_length_ms); |
// Prepare |speech_data_| for encoding, read from a hard-coded file. |
// After preparation, |speech_data_.GetNextBlock()| returns a pointer to a |
@@ -53,6 +53,9 @@ class OpusTest : public TestWithParam<::testing::tuple<int, int>> { |
void SetMaxPlaybackRate(WebRtcOpusEncInst* encoder, |
opus_int32 expect, int32_t set); |
+ void CheckAudioBounded(const int16_t* audio, size_t samples, int channels, |
+ uint16_t bound) const; |
+ |
WebRtcOpusEncInst* opus_encoder_; |
WebRtcOpusDecInst* opus_decoder_; |
@@ -95,6 +98,16 @@ void OpusTest::SetMaxPlaybackRate(WebRtcOpusEncInst* encoder, |
EXPECT_EQ(expect, bandwidth); |
} |
+void OpusTest::CheckAudioBounded(const int16_t* audio, size_t samples, |
+ int channels, uint16_t bound) const { |
+ for (size_t i = 0; i < samples; ++i) { |
+ for (int c = 0; c < channels; ++c) { |
+ ASSERT_GE(audio[i * channels + c], -bound); |
+ ASSERT_LE(audio[i * channels + c], bound); |
+ } |
+ } |
+} |
+ |
int OpusTest::EncodeDecode(WebRtcOpusEncInst* encoder, |
rtc::ArrayView<const int16_t> input_audio, |
WebRtcOpusDecInst* decoder, |
@@ -116,8 +129,9 @@ int OpusTest::EncodeDecode(WebRtcOpusEncInst* encoder, |
// Test if encoder/decoder can enter DTX mode properly and do not enter DTX when |
// they should not. This test is signal dependent. |
-void OpusTest::TestDtxEffect(bool dtx) { |
- PrepareSpeechData(channels_, 20, 2000); |
+void OpusTest::TestDtxEffect(bool dtx, int block_length_ms) { |
+ PrepareSpeechData(channels_, block_length_ms, 2000); |
+ const size_t samples = kOpusRateKhz * block_length_ms; |
// Create encoder memory. |
EXPECT_EQ(0, WebRtcOpus_EncoderCreate(&opus_encoder_, |
@@ -130,17 +144,17 @@ void OpusTest::TestDtxEffect(bool dtx) { |
channels_ == 1 ? 32000 : 64000)); |
// Set input audio as silence. |
- std::vector<int16_t> silence(kOpus20msFrameSamples * channels_, 0); |
+ std::vector<int16_t> silence(samples * channels_, 0); |
// Setting DTX. |
EXPECT_EQ(0, dtx ? WebRtcOpus_EnableDtx(opus_encoder_) : |
WebRtcOpus_DisableDtx(opus_encoder_)); |
int16_t audio_type; |
- int16_t* output_data_decode = new int16_t[kOpus20msFrameSamples * channels_]; |
+ int16_t* output_data_decode = new int16_t[samples * channels_]; |
for (int i = 0; i < 100; ++i) { |
- EXPECT_EQ(kOpus20msFrameSamples, |
+ EXPECT_EQ(samples, |
static_cast<size_t>(EncodeDecode( |
opus_encoder_, speech_data_.GetNextBlock(), opus_decoder_, |
output_data_decode, &audio_type))); |
@@ -157,9 +171,10 @@ void OpusTest::TestDtxEffect(bool dtx) { |
// We input some silent segments. In DTX mode, the encoder will stop sending. |
// However, DTX may happen after a while. |
for (int i = 0; i < 30; ++i) { |
- EXPECT_EQ(kOpus20msFrameSamples, static_cast<size_t>(EncodeDecode( |
- opus_encoder_, silence, opus_decoder_, |
- output_data_decode, &audio_type))); |
+ EXPECT_EQ(samples, |
+ static_cast<size_t>(EncodeDecode( |
+ opus_encoder_, silence, opus_decoder_, output_data_decode, |
+ &audio_type))); |
if (!dtx) { |
EXPECT_GT(encoded_bytes_, 1U); |
EXPECT_EQ(0, opus_encoder_->in_dtx_mode); |
@@ -175,21 +190,47 @@ void OpusTest::TestDtxEffect(bool dtx) { |
// When Opus is in DTX, it wakes up in a regular basis. It sends two packets, |
// one with an arbitrary size and the other of 1-byte, then stops sending for |
- // 19 frames. |
- const int cycles = 5; |
- for (int j = 0; j < cycles; ++j) { |
- // DTX mode is maintained 19 frames. |
- for (int i = 0; i < 19; ++i) { |
- EXPECT_EQ(kOpus20msFrameSamples, |
- static_cast<size_t>( |
- EncodeDecode(opus_encoder_, silence, opus_decoder_, |
- output_data_decode, &audio_type))); |
+ // a certain number of frames. |
+ |
+ // |max_dtx_frames| is the maximum number of frames Opus can stay in DTX. |
+ const int max_dtx_frames = 400 / block_length_ms + 1; |
+ |
+ // We run |kRunTimeMs| milliseconds of pure silence. |
+ const int kRunTimeMs = 2000; |
+ |
+ // We check that, after a |kCheckTimeMs| milliseconds (given that the CNG in |
+ // Opus needs time to adapt), the absolute values of DTX decoded signal are |
+ // bounded by |kOutputValueBound|. |
+ const int kCheckTimeMs = 1500; |
+ |
+#if defined(OPUS_FIXED_POINT) |
+ const uint16_t kOutputValueBound = 20; |
+#else |
+ const uint16_t kOutputValueBound = 2; |
+#endif |
+ |
+ int time = 0; |
+ while (time < kRunTimeMs) { |
+ // DTX mode is maintained for maximum |max_dtx_frames| frames. |
+ int i = 0; |
+ for (; i < max_dtx_frames; ++i) { |
+ time += block_length_ms; |
+ EXPECT_EQ(samples, |
+ static_cast<size_t>(EncodeDecode( |
+ opus_encoder_, silence, opus_decoder_, output_data_decode, |
+ &audio_type))); |
if (dtx) { |
+ if (encoded_bytes_ > 1) |
+ break; |
EXPECT_EQ(0U, encoded_bytes_) // Send 0 byte. |
<< "Opus should have entered DTX mode."; |
EXPECT_EQ(1, opus_encoder_->in_dtx_mode); |
EXPECT_EQ(1, opus_decoder_->in_dtx_mode); |
EXPECT_EQ(2, audio_type); // Comfort noise. |
+ if (time >= kCheckTimeMs) { |
+ CheckAudioBounded(output_data_decode, samples, channels_, |
+ kOutputValueBound); |
+ } |
} else { |
EXPECT_GT(encoded_bytes_, 1U); |
EXPECT_EQ(0, opus_encoder_->in_dtx_mode); |
@@ -198,25 +239,31 @@ void OpusTest::TestDtxEffect(bool dtx) { |
} |
} |
- // Quit DTX after 19 frames. |
- EXPECT_EQ(kOpus20msFrameSamples, static_cast<size_t>(EncodeDecode( |
- opus_encoder_, silence, opus_decoder_, |
- output_data_decode, &audio_type))); |
+ if (dtx) { |
+ // With DTX, Opus must stop transmission for some time. |
+ EXPECT_GT(i, 1); |
+ } |
- EXPECT_GT(encoded_bytes_, 1U); |
+ // We expect a normal payload. |
EXPECT_EQ(0, opus_encoder_->in_dtx_mode); |
EXPECT_EQ(0, opus_decoder_->in_dtx_mode); |
EXPECT_EQ(0, audio_type); // Speech. |
// Enters DTX again immediately. |
- EXPECT_EQ(kOpus20msFrameSamples, static_cast<size_t>(EncodeDecode( |
- opus_encoder_, silence, opus_decoder_, |
- output_data_decode, &audio_type))); |
+ time += block_length_ms; |
+ EXPECT_EQ(samples, |
+ static_cast<size_t>(EncodeDecode( |
+ opus_encoder_, silence, opus_decoder_, output_data_decode, |
+ &audio_type))); |
if (dtx) { |
EXPECT_EQ(1U, encoded_bytes_); // Send 1 byte. |
EXPECT_EQ(1, opus_encoder_->in_dtx_mode); |
EXPECT_EQ(1, opus_decoder_->in_dtx_mode); |
EXPECT_EQ(2, audio_type); // Comfort noise. |
+ if (time >= kCheckTimeMs) { |
+ CheckAudioBounded(output_data_decode, samples, channels_, |
+ kOutputValueBound); |
+ } |
} else { |
EXPECT_GT(encoded_bytes_, 1U); |
EXPECT_EQ(0, opus_encoder_->in_dtx_mode); |
@@ -228,9 +275,10 @@ void OpusTest::TestDtxEffect(bool dtx) { |
silence[0] = 10000; |
if (dtx) { |
// Verify that encoder/decoder can jump out from DTX mode. |
- EXPECT_EQ(kOpus20msFrameSamples, static_cast<size_t>(EncodeDecode( |
- opus_encoder_, silence, opus_decoder_, |
- output_data_decode, &audio_type))); |
+ EXPECT_EQ(samples, |
+ static_cast<size_t>(EncodeDecode( |
+ opus_encoder_, silence, opus_decoder_, output_data_decode, |
+ &audio_type))); |
EXPECT_GT(encoded_bytes_, 1U); |
EXPECT_EQ(0, opus_encoder_->in_dtx_mode); |
EXPECT_EQ(0, opus_decoder_->in_dtx_mode); |
@@ -436,11 +484,15 @@ TEST_P(OpusTest, OpusEnableDisableDtx) { |
} |
TEST_P(OpusTest, OpusDtxOff) { |
- TestDtxEffect(false); |
+ TestDtxEffect(false, 10); |
+ TestDtxEffect(false, 20); |
+ TestDtxEffect(false, 40); |
} |
TEST_P(OpusTest, OpusDtxOn) { |
- TestDtxEffect(true); |
+ TestDtxEffect(true, 10); |
+ TestDtxEffect(true, 20); |
+ TestDtxEffect(true, 40); |
} |
TEST_P(OpusTest, OpusSetPacketLossRate) { |