tune speaker detector more sensitive (#427)

* tune speaker detector more sensitive

pkg/config/config.go

@@ -192,9 +192,9 @@ func NewConfig(confString string, c *cli.Context) (*Config, error) {
 			},
 		},
 		Audio: AudioConfig{
-			ActiveLevel:     30, // -30dBov = 0.03
+			ActiveLevel:     35, // -35dBov
 			MinPercentile:   40,
-			UpdateInterval:  500,
+			UpdateInterval:  400,
 			SmoothIntervals: 2,
 		},
 		Redis: RedisConfig{},

pkg/rtc/audiolevel.go

@@ -7,7 +7,6 @@ import (
 
 const (
 	// duration of audio frames for observe window
-	observeDuration  = 500 // ms
 	SilentAudioLevel = 127
 )
 
@@ -20,17 +19,19 @@ type AudioLevel struct {
 
 	// for Observe goroutine use
 	// keeps track of current activity
-	observeLevel     uint8
-	activeDuration   uint32 // ms
-	observedDuration uint32 // ms
+	observeLevel      uint8
+	activeDuration    uint32 // ms
+	observedDuration  uint32 // ms
+	durationToObserve uint32 // ms
 }
 
-func NewAudioLevel(activeLevel uint8, minPercentile uint8) *AudioLevel {
+func NewAudioLevel(activeLevel uint8, minPercentile uint8, observeDuration uint32) *AudioLevel {
 	l := &AudioLevel{
 		levelThreshold:    activeLevel,
 		minActiveDuration: uint32(minPercentile) * observeDuration / 100,
 		currentLevel:      SilentAudioLevel,
 		observeLevel:      SilentAudioLevel,
+		durationToObserve: observeDuration,
 	}
 	return l
 }
@@ -46,10 +47,10 @@ func (l *AudioLevel) Observe(level uint8, durationMs uint32) {
 		}
 	}
 
-	if l.observedDuration >= observeDuration {
+	if l.observedDuration >= l.durationToObserve {
 		// compute and reset
 		if l.activeDuration >= l.minActiveDuration {
-			level := uint32(l.observeLevel) - uint32(20*math.Log10(float64(l.activeDuration)/float64(observeDuration)))
+			level := uint32(l.observeLevel) - uint32(20*math.Log10(float64(l.activeDuration)/float64(l.durationToObserve)))
 			atomic.StoreUint32(&l.currentLevel, level)
 		} else {
 			atomic.StoreUint32(&l.currentLevel, SilentAudioLevel)

pkg/rtc/audiolevel_test.go

@@ -12,12 +12,13 @@ const (
 	samplesPerBatch    = 25
 	defaultActiveLevel = 30
 	// requires two noisy samples to count
-	defaultPercentile = 10
+	defaultPercentile      = 10
+	defaultObserveDuration = 500 // ms
 )
 
 func TestAudioLevel(t *testing.T) {
 	t.Run("initially to return not noisy, within a few samples", func(t *testing.T) {
-		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile)
+		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile, defaultObserveDuration)
 		_, noisy := a.GetLevel()
 		require.False(t, noisy)
 
@@ -27,7 +28,7 @@ func TestAudioLevel(t *testing.T) {
 	})
 
 	t.Run("not noisy when all samples are below threshold", func(t *testing.T) {
-		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile)
+		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile, defaultObserveDuration)
 
 		observeSamples(a, 35, 100)
 		_, noisy := a.GetLevel()
@@ -35,7 +36,7 @@ func TestAudioLevel(t *testing.T) {
 	})
 
 	t.Run("not noisy when less than percentile samples are above threshold", func(t *testing.T) {
-		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile)
+		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile, defaultObserveDuration)
 
 		observeSamples(a, 35, samplesPerBatch-2)
 		observeSamples(a, 25, 1)
@@ -46,7 +47,7 @@ func TestAudioLevel(t *testing.T) {
 	})
 
 	t.Run("noisy when higher than percentile samples are above threshold", func(t *testing.T) {
-		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile)
+		a := rtc.NewAudioLevel(defaultActiveLevel, defaultPercentile, defaultObserveDuration)
 
 		observeSamples(a, 35, samplesPerBatch-16)
 		observeSamples(a, 25, 8)

pkg/rtc/mediatrack.go

@@ -125,7 +125,7 @@ func (t *MediaTrack) AddReceiver(receiver *webrtc.RTPReceiver, track *webrtc.Tra
 
 	if t.Kind() == livekit.TrackType_AUDIO {
 		t.audioLevelMu.Lock()
-		t.audioLevel = NewAudioLevel(t.params.AudioConfig.ActiveLevel, t.params.AudioConfig.MinPercentile)
+		t.audioLevel = NewAudioLevel(t.params.AudioConfig.ActiveLevel, t.params.AudioConfig.MinPercentile, t.params.AudioConfig.UpdateInterval)
 		buff.OnAudioLevel(func(level uint8, duration uint32) {
 			t.audioLevelMu.RLock()
 			defer t.audioLevelMu.RUnlock()