Android/smart_human
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

添加asr校验有效人声

Browse Source
This commit is contained in:
林若思 2026-01-04 18:57:44 +08:00
parent aee1d6b797
commit 3143f4007e
1 changed files with 30 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
app/src/main/java/com/zs/smarthuman/sherpa/VoiceController.kt
View File

@ -53,7 +53,8 @@ class VoiceController(
private val idleTimeoutMs = idleTimeoutSeconds * 1000L private val idleTimeoutMs = idleTimeoutSeconds * 1000L
private val maxRecordingMs = maxRecordingSeconds * 1000L private val maxRecordingMs = maxRecordingSeconds * 1000L
private val vadManager = VadManager(assetManager, private val vadManager = VadManager(
assetManager,
onSpeechStart = { onVadStart() }, onSpeechStart = { onVadStart() },
onSpeechEnd = { avgEnergy, peakRms -> onVadEnd(avgEnergy, peakRms) } onSpeechEnd = { avgEnergy, peakRms -> onVadEnd(avgEnergy, peakRms) }
) )
@ -100,7 +101,7 @@ class VoiceController(
val result = it.getResult(stream) val result = it.getResult(stream)
stream.release() stream.release()
result.text result.text
}?:"" } ?: ""
} catch (e: Exception) { } catch (e: Exception) {
Log.e(TAG, "Error during ASR processing: ${e.message}") Log.e(TAG, "Error during ASR processing: ${e.message}")
@ -137,7 +138,8 @@ class VoiceController(
VoiceState.WAIT_SPEECH -> { VoiceState.WAIT_SPEECH -> {
if ((waitSpeechStartMs > 0 && now - waitSpeechStartMs >= idleTimeoutMs) || if ((waitSpeechStartMs > 0 && now - waitSpeechStartMs >= idleTimeoutMs) ||
(waitSpeechFailStartMs > 0 && now - waitSpeechFailStartMs >= idleTimeoutMs)) { (waitSpeechFailStartMs > 0 && now - waitSpeechFailStartMs >= idleTimeoutMs)
) {
Log.d(TAG, "⏱ WAIT_SPEECH idle timeout → WAIT_WAKEUP") Log.d(TAG, "⏱ WAIT_SPEECH idle timeout → WAIT_WAKEUP")
resetAll() resetAll()
return return
@ -215,7 +217,10 @@ class VoiceController(
val peakAvgRatio = if (avgEnergy > 0f) peakRms / avgEnergy else 0f val peakAvgRatio = if (avgEnergy > 0f) peakRms / avgEnergy else 0f
Log.d(TAG, "📊 Finish Sentence - duration: $duration ms, vadEnded: true") Log.d(TAG, "📊 Finish Sentence - duration: $duration ms, vadEnded: true")
Log.d(TAG, "📊 vadRatio=$vadRatio, avgEnergy=$avgEnergy, peakRms=$peakRms, peakAvgRatio=$peakAvgRatio") Log.d(
TAG,
"📊 vadRatio=$vadRatio, avgEnergy=$avgEnergy, peakRms=$peakRms, peakAvgRatio=$peakAvgRatio"
)
if (avgEnergy < 0.02f || peakAvgRatio < 1.2f || vadRatio < 0.4f) { if (avgEnergy < 0.02f || peakAvgRatio < 1.2f || vadRatio < 0.4f) {
Log.d(TAG, "❌ Sentence rejected") Log.d(TAG, "❌ Sentence rejected")
@ -230,7 +235,8 @@ class VoiceController(
return return
} }
val asrText = runSecondPass(audio) var asrText = runSecondPass(audio)
if (asrText.isEmpty()) { if (asrText.isEmpty()) {
resetToWaitSpeech() resetToWaitSpeech()
Log.d(TAG, "识别不出来asr") Log.d(TAG, "识别不出来asr")
@ -256,7 +262,10 @@ class VoiceController(
vadRatio >= 0.55f -> score += 2 vadRatio >= 0.55f -> score += 2
vadRatio >= 0.40f -> score += 1 vadRatio >= 0.40f -> score += 1
} }
Log.d(TAG, "📊 duration=$duration ms, vadRatio=$vadRatio, avgEnergy=$avgEnergy, score=$score") Log.d(
TAG,
"📊 duration=$duration ms, vadRatio=$vadRatio, avgEnergy=$avgEnergy, score=$score"
)
val pass = score >= 5 || (score == 3 && avgEnergy >= 0.06f) val pass = score >= 5 || (score == 3 && avgEnergy >= 0.06f)
if (!pass) { if (!pass) {
@ -271,13 +280,19 @@ class VoiceController(
} }
/* ================= 播放回调 ================= */ /* ================= 播放回调 ================= */
fun onPlayStartPrompt() { state = VoiceState.PLAYING_PROMPT } fun onPlayStartPrompt() {
state = VoiceState.PLAYING_PROMPT
}
fun onPlayEndPrompt() { fun onPlayEndPrompt() {
speechEnableAtMs = System.currentTimeMillis() + SPEECH_COOLDOWN_MS speechEnableAtMs = System.currentTimeMillis() + SPEECH_COOLDOWN_MS
state = VoiceState.WAIT_SPEECH_COOLDOWN state = VoiceState.WAIT_SPEECH_COOLDOWN
} }
fun onPlayStartBackend() { state = VoiceState.PLAYING_BACKEND } fun onPlayStartBackend() {
state = VoiceState.PLAYING_BACKEND
}
fun onPlayEndBackend() { fun onPlayEndBackend() {
speechEnableAtMs = System.currentTimeMillis() + SPEECH_COOLDOWN_MS speechEnableAtMs = System.currentTimeMillis() + SPEECH_COOLDOWN_MS
state = VoiceState.WAIT_SPEECH_COOLDOWN state = VoiceState.WAIT_SPEECH_COOLDOWN