Event logs - separate audio_level and voice_activity

Bug: webrtc:8111
Change-Id: I44d81c5b4f5b854e8accd84521fbbd7b50228903
Reviewed-on: https://webrtc-review.googlesource.com/c/109571
Commit-Queue: Elad Alon <eladalon@webrtc.org>
Reviewed-by: Björn Terelius <terelius@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25589}

logging/rtc_event_log/encoder/rtc_event_log_encoder_new_format.cc

@@ -239,13 +239,6 @@ rtclog2::IceCandidatePairEvent::IceCandidatePairEventType ConvertToProtoFormat(
   return rtclog2::IceCandidatePairEvent::UNKNOWN_CHECK_TYPE;
 }
 
-uint8_t ConvertAudioLevelToProtoFormat(bool voice_activity,
-                                       uint8_t audio_level) {
-  RTC_DCHECK_EQ(audio_level & static_cast<uint8_t>(0x80), 0);
-  constexpr uint8_t kVoiceActivityBit = 0x80;
-  return audio_level | (voice_activity ? kVoiceActivityBit : 0);
-}
-
 // Copies all RTCP blocks except APP, SDES and unknown from |packet| to
 // |buffer|. |buffer| must have space for |IP_PACKET_SIZE| bytes. |packet| must
 // be at most |IP_PACKET_SIZE| bytes long.
@@ -338,8 +331,10 @@ void EncodeRtcpPacket(rtc::ArrayView<const EventType*> batch,
     static_assert(sizeof(std::string::value_type) == sizeof(uint8_t), "");
     const size_t buffer_length = RemoveNonWhitelistedRtcpBlocks(
         event->packet_, reinterpret_cast<uint8_t*>(&scrubed_packets[i][0]));
+    if (buffer_length < event->packet_.size()) {
       scrubed_packets[i].resize(buffer_length);
     }
+  }
   proto_batch->set_raw_packet_blobs(EncodeBlobs(scrubed_packets));
 }
 
@@ -405,17 +400,19 @@ void EncodeRtpPacket(const std::vector<const EventType*>& batch,
     }
   }
 
-  // TODO(eladalon): Separate audio level from voice activity.
   absl::optional<uint64_t> base_audio_level;
+  absl::optional<uint64_t> base_voice_activity;
   {
     bool voice_activity;
     uint8_t audio_level;
     if (base_event->header_.template GetExtension<AudioLevel>(&voice_activity,
                                                               &audio_level)) {
-      proto_batch->set_audio_level(
+      RTC_DCHECK_LE(audio_level, 0x7Fu);
-          ConvertAudioLevelToProtoFormat(voice_activity, audio_level));
+      base_audio_level = audio_level;
-      base_audio_level =
+      proto_batch->set_audio_level(audio_level);
-          ConvertAudioLevelToProtoFormat(voice_activity, audio_level);
+
+      base_voice_activity = voice_activity;
+      proto_batch->set_voice_activity(voice_activity);
     }
   }
 
@@ -587,7 +584,8 @@ void EncodeRtpPacket(const std::vector<const EventType*>& batch,
     uint8_t audio_level;
     if (event->header_.template GetExtension<AudioLevel>(&voice_activity,
                                                          &audio_level)) {
-      values[i] = ConvertAudioLevelToProtoFormat(voice_activity, audio_level);
+      RTC_DCHECK_LE(audio_level, 0x7Fu);
+      values[i] = audio_level;
     } else {
       values[i].reset();
     }
@@ -596,6 +594,24 @@ void EncodeRtpPacket(const std::vector<const EventType*>& batch,
   if (!encoded_deltas.empty()) {
     proto_batch->set_audio_level_deltas(encoded_deltas);
   }
+
+  // voice_activity (RTP extension)
+  for (size_t i = 0; i < values.size(); ++i) {
+    const EventType* event = batch[i + 1];
+    bool voice_activity;
+    uint8_t audio_level;
+    if (event->header_.template GetExtension<AudioLevel>(&voice_activity,
+                                                         &audio_level)) {
+      RTC_DCHECK_LE(audio_level, 0x7Fu);
+      values[i] = voice_activity;
+    } else {
+      values[i].reset();
+    }
+  }
+  encoded_deltas = EncodeDeltas(base_voice_activity, values);
+  if (!encoded_deltas.empty()) {
+    proto_batch->set_voice_activity_deltas(encoded_deltas);
+  }
 }
 }  // namespace
 

logging/rtc_event_log/rtc_event_log2.proto

@@ -86,7 +86,9 @@ message IncomingRtpPackets {
   optional int32 transmission_time_offset = 16;
   optional uint32 absolute_send_time = 17;
   optional uint32 video_rotation = 18;
+  // |audio_level| and |voice_activity| are always used in conjunction.
   optional uint32 audio_level = 19;
+  optional bool voice_activity = 20;
   // TODO(terelius): Add other header extensions like playout delay?
 
   // Delta encodings.
@@ -105,7 +107,9 @@ message IncomingRtpPackets {
   optional bytes transmission_time_offset_deltas = 116;
   optional bytes absolute_send_time_deltas = 117;
   optional bytes video_rotation_deltas = 118;
+  // |audio_level| and |voice_activity| are always used in conjunction.
   optional bytes audio_level_deltas = 119;
+  optional bytes voice_activity_deltas = 120;
 }
 
 message OutgoingRtpPackets {
@@ -150,7 +154,9 @@ message OutgoingRtpPackets {
   optional int32 transmission_time_offset = 16;
   optional uint32 absolute_send_time = 17;
   optional uint32 video_rotation = 18;
+  // |audio_level| and |voice_activity| are always used in conjunction.
   optional uint32 audio_level = 19;
+  optional bool voice_activity = 20;
   // TODO(terelius): Add other header extensions like playout delay?
 
   // Delta encodings.
@@ -169,7 +175,9 @@ message OutgoingRtpPackets {
   optional bytes transmission_time_offset_deltas = 116;
   optional bytes absolute_send_time_deltas = 117;
   optional bytes video_rotation_deltas = 118;
+  // |audio_level| and |voice_activity| are always used in conjunction.
   optional bytes audio_level_deltas = 119;
+  optional bytes voice_activity_deltas = 120;
 }
 
 message IncomingRtcpPackets {

logging/rtc_event_log/rtc_event_log_parser_new.cc

@@ -497,11 +497,16 @@ void StoreRtpPackets(
           rtc::checked_cast<uint8_t>(proto.video_rotation()));
     }
     if (proto.has_audio_level()) {
+      RTC_CHECK(proto.has_voice_activity());
       header.extension.hasAudioLevel = true;
+      header.extension.voiceActivity =
+          rtc::checked_cast<bool>(proto.voice_activity());
       const uint8_t audio_level =
           rtc::checked_cast<uint8_t>(proto.audio_level());
-      header.extension.voiceActivity = (audio_level >> 7) != 0;
+      RTC_CHECK_LE(audio_level, 0x7Fu);
-      header.extension.audioLevel = audio_level & 0x7Fu;
+      header.extension.audioLevel = audio_level;
+    } else {
+      RTC_CHECK(!proto.has_voice_activity());
     }
     (*rtp_packets_map)[header.ssrc].emplace_back(
         proto.timestamp_ms() * 1000, header, proto.header_size(),
@@ -620,6 +625,17 @@ void StoreRtpPackets(
     RTC_CHECK_EQ(audio_level_values.size(), number_of_deltas);
   }
 
+  // voice_activity (RTP extension)
+  std::vector<absl::optional<uint64_t>> voice_activity_values;
+  {
+    const absl::optional<uint64_t> base_voice_activity =
+        proto.has_voice_activity() ? proto.voice_activity()
+                                   : absl::optional<uint64_t>();
+    voice_activity_values = DecodeDeltas(proto.voice_activity_deltas(),
+                                         base_voice_activity, number_of_deltas);
+    RTC_CHECK_EQ(voice_activity_values.size(), number_of_deltas);
+  }
+
   // Delta decoding
   for (size_t i = 0; i < number_of_deltas; ++i) {
     RTC_CHECK(timestamp_ms_values[i].has_value());
@@ -670,11 +686,18 @@ void StoreRtpPackets(
           rtc::checked_cast<uint8_t>(video_rotation_values[i].value()));
     }
     if (audio_level_values.size() > i && audio_level_values[i].has_value()) {
+      RTC_CHECK(voice_activity_values.size() > i &&
+                voice_activity_values[i].has_value());
       header.extension.hasAudioLevel = true;
+      header.extension.voiceActivity =
+          rtc::checked_cast<bool>(voice_activity_values[i].value());
       const uint8_t audio_level =
           rtc::checked_cast<uint8_t>(audio_level_values[i].value());
-      header.extension.voiceActivity = (audio_level >> 7) != 0;
+      RTC_CHECK_LE(audio_level, 0x7Fu);
-      header.extension.audioLevel = audio_level & 0x7Fu;
+      header.extension.audioLevel = audio_level;
+    } else {
+      RTC_CHECK(voice_activity_values.size() <= i ||
+                !voice_activity_values[i].has_value());
     }
     (*rtp_packets_map)[header.ssrc].emplace_back(
         timestamp_ms_values[i].value() * 1000, header, header.headerLength,
@@ -2300,6 +2323,7 @@ void ParsedRtcEventLogNew::StoreAudioNetworkAdaptationEvent(
       runtime_config.enable_dtx = proto.enable_dtx();
     }
     if (proto.has_num_channels()) {
+      // TODO(eladalon): Encode 1/2 -> 0/1, to improve
       runtime_config.num_channels = proto.num_channels();
     }
     audio_network_adaptation_events_.emplace_back(1000 * proto.timestamp_ms(),