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

opus: take SILK vad result into account for voice detection

Browse Source 
BUG=webrtc:11643

Change-Id: Idc3a9b6bb7bd1a33f905843e5d6067ae19d5172c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176508
Commit-Queue: Minyue Li <minyue@webrtc.org>
Reviewed-by: Minyue Li <minyue@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31743}
This commit is contained in:
Philipp Hancke
2020-07-16 09:47:24 +02:00
committed by Commit Bot
parent 3592839896
commit 686a3709ac
5 changed files with 60 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
modules/audio_coding/codecs/opus/audio_encoder_opus.cc
View File

@ -367,8 +367,7 @@ AudioEncoderOpusImpl::AudioEncoderOpusImpl(
inst_(nullptr),
packet_loss_fraction_smoother_(new PacketLossFractionSmoother()),
audio_network_adaptor_creator_(audio_network_adaptor_creator),
bitrate_smoother_(std::move(bitrate_smoother)),
consecutive_dtx_frames_(0) {
bitrate_smoother_(std::move(bitrate_smoother)) {
RTC_DCHECK(0 <= payload_type && payload_type <= 127);
// Sanity check of the redundant payload type field that we want to get rid
@ -590,6 +589,7 @@ AudioEncoder::EncodedInfo AudioEncoderOpusImpl::EncodeImpl(
Num10msFramesPerPacket() * SamplesPer10msFrame());
const size_t max_encoded_bytes = SufficientOutputBufferSize();
const size_t start_offset_bytes = encoded->size();
EncodedInfo info;
info.encoded_bytes = encoded->AppendData(
max_encoded_bytes, [&](rtc::ArrayView<uint8_t> encoded) {
@ -604,8 +604,6 @@ AudioEncoder::EncodedInfo AudioEncoderOpusImpl::EncodeImpl(
});
input_buffer_.clear();
bool dtx_frame = (info.encoded_bytes <= 2);
// Will use new packet size for next encoding.
config_.frame_size_ms = next_frame_length_ms_;
@ -620,14 +618,18 @@ AudioEncoder::EncodedInfo AudioEncoderOpusImpl::EncodeImpl(
info.encoded_timestamp = first_timestamp_in_buffer_;
info.payload_type = payload_type_;
info.send_even_if_empty = true; // Allows Opus to send empty packets.
// After 20 DTX frames (MAX_CONSECUTIVE_DTX) Opus will send a frame
// coding the background noise. Avoid flagging this frame as speech
// (even though there is a probability of the frame being speech).
info.speech = !dtx_frame && (consecutive_dtx_frames_ != 20);
info.encoder_type = CodecType::kOpus;
// Increase or reset DTX counter.
consecutive_dtx_frames_ = (dtx_frame) ? (consecutive_dtx_frames_ + 1) : (0);
// Extract the VAD result from the encoded packet.
int has_voice = WebRtcOpus_PacketHasVoiceActivity(
&encoded->data()[start_offset_bytes], info.encoded_bytes);
if (has_voice == -1) {
// CELT mode packet or there was an error. This had set the speech flag to
// true historically.
info.speech = true;
} else {
info.speech = has_voice;
}
return info;
}

1
modules/audio_coding/codecs/opus/audio_encoder_opus.h
View File

@ -172,7 +172,6 @@ class AudioEncoderOpusImpl final : public AudioEncoder {
absl::optional<size_t> overhead_bytes_per_packet_;
const std::unique_ptr<SmoothingFilter> bitrate_smoother_;
absl::optional<int64_t> bitrate_smoother_last_update_time_;
int consecutive_dtx_frames_;
friend struct AudioEncoderOpus;
RTC_DISALLOW_COPY_AND_ASSIGN(AudioEncoderOpusImpl);

2
modules/audio_coding/codecs/opus/opus_interface.cc
View File

@ -767,7 +767,7 @@ int WebRtcOpus_PacketHasVoiceActivity(const uint8_t* payload,
int silk_frames = WebRtcOpus_NumSilkFrames(payload);
if (silk_frames == 0)
return -1;
return 0;
const int channels = opus_packet_get_nb_channels(payload);
RTC_DCHECK(channels == 1 || channels == 2);

17
modules/audio_coding/codecs/opus/opus_unittest.cc
View File

@ -975,4 +975,21 @@ TEST(OpusVadTest, TwoOpusMonoFramesVadOnSecond) {
EXPECT_TRUE(WebRtcOpus_PacketHasVoiceActivity(twoMonoFrames, 3));
}
TEST(OpusVadTest, DtxEmptyPacket) {
const uint8_t dtx[] = {0x78};
EXPECT_FALSE(WebRtcOpus_PacketHasVoiceActivity(dtx, 1));
}
TEST(OpusVadTest, DtxBackgroundNoisePacket) {
// DTX sends a frame coding background noise every 20 packets:
// https://tools.ietf.org/html/rfc6716#section-2.1.9
// The packet below represents such a frame and was captured using
// Wireshark while disabling encryption.
const uint8_t dtx[] = {0x78, 0x07, 0xc9, 0x79, 0xc8, 0xc9, 0x57, 0xc0, 0xa2,
0x12, 0x23, 0xfa, 0xef, 0x67, 0xf3, 0x2e, 0xe3, 0xd3,
0xd5, 0xe9, 0xec, 0xdb, 0x3e, 0xbc, 0x80, 0xb6, 0x6e,
0x2a, 0xb7, 0x8c, 0x83, 0xcd, 0x83, 0xcd, 0x00};
EXPECT_FALSE(WebRtcOpus_PacketHasVoiceActivity(dtx, 35));
}
} // namespace webrtc