Fix issue #159 (#195)

* Fix issue #159

use timestamp for AudeioLevel observer
change smoothinterval default to 2 for more sensitive

pkg/config/config.go

@@ -146,7 +146,7 @@ func NewConfig(confString string, c *cli.Context) (*Config, error) {
 			ActiveLevel:     30, // -30dBov = 0.03
 			MinPercentile:   40,
 			UpdateInterval:  500,
-			SmoothIntervals: 4,
+			SmoothIntervals: 2,
 		},
 		Redis: RedisConfig{},
 		Room: RoomConfig{

pkg/rtc/audiolevel.go

@@ -6,8 +6,8 @@ import (
 )
 
 const (
-	// number of audio frames for observe window
-	observeFrames    = 25 // webrtc default opus frame size 20ms, 25*20=500ms matching default UpdateInterval
+	// duration of audio frames for observe window
+	observeDuration  = 500 // ms
 	silentAudioLevel = 127
 )
 
@@ -15,49 +15,48 @@ const (
 type AudioLevel struct {
 	levelThreshold uint8
 	currentLevel   uint32
-	// min frames to be considered active
-	minActiveFrames uint32
+	// min duration to be considered active
+	minActiveDuration uint32
 
 	// for Observe goroutine use
 	// keeps track of current activity
-	observeLevel uint8
-	activeFrames uint32
-	numFrames    uint32
+	observeLevel     uint8
+	activeDuration   uint32 // ms
+	observedDuration uint32 // ms
 }
 
 func NewAudioLevel(activeLevel uint8, minPercentile uint8) *AudioLevel {
 	l := &AudioLevel{
-		levelThreshold:  activeLevel,
-		minActiveFrames: uint32(minPercentile) * observeFrames / 100,
-		currentLevel:    silentAudioLevel,
-		observeLevel:    silentAudioLevel,
+		levelThreshold:    activeLevel,
+		minActiveDuration: uint32(minPercentile) * observeDuration / 100,
+		currentLevel:      silentAudioLevel,
+		observeLevel:      silentAudioLevel,
 	}
 	return l
 }
 
 // Observes a new frame, must be called from the same thread
-func (l *AudioLevel) Observe(level uint8) {
-	l.numFrames++
+func (l *AudioLevel) Observe(level uint8, durationMs uint32) {
+	l.observedDuration += durationMs
 
 	if level <= l.levelThreshold {
-		l.activeFrames++
+		l.activeDuration += durationMs
 		if l.observeLevel > level {
 			l.observeLevel = level
 		}
 	}
 
-	if l.numFrames >= observeFrames {
+	if l.observedDuration >= observeDuration {
 		// compute and reset
-		if l.activeFrames >= l.minActiveFrames {
-			const invObserveFrames = 1.0 / observeFrames
-			level := uint32(l.observeLevel) - uint32(20*math.Log10(float64(l.activeFrames)*invObserveFrames))
+		if l.activeDuration >= l.minActiveDuration {
+			level := uint32(l.observeLevel) - uint32(20*math.Log10(float64(l.activeDuration)/float64(observeDuration)))
 			atomic.StoreUint32(&l.currentLevel, level)
 		} else {
 			atomic.StoreUint32(&l.currentLevel, silentAudioLevel)
 		}
 		l.observeLevel = silentAudioLevel
-		l.activeFrames = 0
-		l.numFrames = 0
+		l.activeDuration = 0
+		l.observedDuration = 0
 	}
 }
 

pkg/rtc/audiolevel_test.go

@@ -60,6 +60,6 @@ func TestAudioLevel(t *testing.T) {
 
 func observeSamples(a *rtc.AudioLevel, level uint8, count int) {
 	for i := 0; i < count; i++ {
-		a.Observe(level)
+		a.Observe(level, 20)
 	}
 }

pkg/rtc/mediatrack.go

@@ -344,8 +344,8 @@ func (t *MediaTrack) AddReceiver(receiver *webrtc.RTPReceiver, track *webrtc.Tra
 
 	if t.Kind() == livekit.TrackType_AUDIO {
 		t.audioLevel = NewAudioLevel(t.params.AudioConfig.ActiveLevel, t.params.AudioConfig.MinPercentile)
-		buff.OnAudioLevel(func(level uint8) {
-			t.audioLevel.Observe(level)
+		buff.OnAudioLevel(func(level uint8, duration uint32) {
+			t.audioLevel.Observe(level, duration)
 		})
 	} else if t.Kind() == livekit.TrackType_VIDEO {
 		if twcc != nil {

pkg/sfu/buffer/buffer.go

@@ -88,7 +88,7 @@ type Buffer struct {
 
 	// callbacks
 	onClose      func()
-	onAudioLevel func(level uint8)
+	onAudioLevel func(level uint8, durationMs uint32)
 	feedbackCB   func([]rtcp.Packet)
 	feedbackTWCC func(sn uint16, timeNS int64, marker bool)
 
@@ -386,7 +386,10 @@ func (b *Buffer) calc(pkt []byte, arrivalTime int64) {
 		if e := p.GetExtension(b.audioExt); e != nil && b.onAudioLevel != nil {
 			ext := rtp.AudioLevelExtension{}
 			if err := ext.Unmarshal(e); err == nil {
-				b.onAudioLevel(ext.Level)
+				duration := (int64(p.Timestamp) - int64(latestTimestamp)) * 1e3 / int64(b.clockRate)
+				if duration > 0 {
+					b.onAudioLevel(ext.Level, uint32(duration))
+				}
 			}
 		}
 	}
@@ -607,7 +610,7 @@ func (b *Buffer) OnFeedback(fn func(fb []rtcp.Packet)) {
 	b.feedbackCB = fn
 }
 
-func (b *Buffer) OnAudioLevel(fn func(level uint8)) {
+func (b *Buffer) OnAudioLevel(fn func(level uint8, durationMs uint32)) {
 	b.onAudioLevel = fn
 }