暂时去除asr模型，部分机器带不动
This commit is contained in:
parent
a1d0520789
commit
8dd5e68ded
Binary file not shown.
@ -1,4 +0,0 @@
|
||||
# Introduction
|
||||
|
||||
Model in this directory is converted from
|
||||
https://huggingface.co/ASLP-lab/WSYue-ASR/tree/main/u2pp_conformer_yue
|
||||
Binary file not shown.
File diff suppressed because it is too large
Load Diff
@ -2,17 +2,12 @@ package com.zs.smarthuman.sherpa
|
||||
|
||||
import android.content.res.AssetManager
|
||||
import android.util.Log
|
||||
import com.k2fsa.sherpa.onnx.OfflineModelConfig
|
||||
import com.k2fsa.sherpa.onnx.OfflineRecognizer
|
||||
import com.k2fsa.sherpa.onnx.OfflineRecognizerConfig
|
||||
import com.k2fsa.sherpa.onnx.OfflineWenetCtcModelConfig
|
||||
import com.k2fsa.sherpa.onnx.getFeatureConfig
|
||||
import java.util.ArrayDeque
|
||||
|
||||
class VoiceController(
|
||||
assetManager: AssetManager,
|
||||
private val onWakeup: () -> Unit,
|
||||
private val onFinalAudio: (FloatArray, String) -> Unit, // 修改：传回识别结果文本
|
||||
private val onFinalAudio: (FloatArray) -> Unit, // 修改：传回识别结果文本
|
||||
private val idleTimeoutSeconds: Int = 10,
|
||||
private val maxRecordingSeconds: Int = 10,
|
||||
private val onStateChanged: ((VoiceState) -> Unit)? = null,
|
||||
@ -59,57 +54,6 @@ class VoiceController(
|
||||
onSpeechEnd = { avgEnergy, peakRms -> onVadEnd(avgEnergy, peakRms) }
|
||||
)
|
||||
|
||||
|
||||
// 初始化非流式 ASR
|
||||
private var offlineRecognizer: OfflineRecognizer? = null
|
||||
|
||||
init {
|
||||
offlineRecognizer = initOfflineRecognizer(assetManager)
|
||||
}
|
||||
|
||||
private fun initOfflineRecognizer(assetManager: AssetManager): OfflineRecognizer {
|
||||
// Wenetspeech 模型配置
|
||||
val wenetConfig = OfflineWenetCtcModelConfig(
|
||||
model = "sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10/model.int8.onnx"
|
||||
)
|
||||
|
||||
val modelConfig = OfflineModelConfig(
|
||||
wenetCtc = wenetConfig,
|
||||
tokens = "sherpa-onnx-wenetspeech-yue-u2pp-conformer-ctc-zh-en-cantonese-int8-2025-09-10/tokens.txt",
|
||||
numThreads = 4,
|
||||
debug = true
|
||||
)
|
||||
|
||||
val featConfig = getFeatureConfig(sampleRate = sampleRate, featureDim = 80)
|
||||
|
||||
return OfflineRecognizer(
|
||||
assetManager = assetManager,
|
||||
config = OfflineRecognizerConfig(
|
||||
featConfig = featConfig,
|
||||
modelConfig = modelConfig
|
||||
)
|
||||
)
|
||||
}
|
||||
|
||||
// runSecondPass 里直接用 offlineRecognizer
|
||||
private fun runSecondPass(samples: FloatArray): String {
|
||||
return try {
|
||||
offlineRecognizer?.let {
|
||||
val stream = it.createStream()
|
||||
stream.acceptWaveform(samples, sampleRate)
|
||||
it.decode(stream)
|
||||
val result = it.getResult(stream)
|
||||
stream.release()
|
||||
result.text
|
||||
} ?: ""
|
||||
|
||||
} catch (e: Exception) {
|
||||
Log.e(TAG, "Error during ASR processing: ${e.message}")
|
||||
""
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
/* ================= 音频入口 ================= */
|
||||
fun acceptAudio(samples: FloatArray) {
|
||||
cachePreBuffer(samples)
|
||||
@ -228,24 +172,6 @@ class VoiceController(
|
||||
return
|
||||
}
|
||||
|
||||
// 检查音频数据是否有效
|
||||
if (audio.isEmpty()) {
|
||||
Log.d(TAG, "❌ Audio is empty")
|
||||
resetToWaitSpeech()
|
||||
return
|
||||
}
|
||||
|
||||
var asrText = runSecondPass(audio)
|
||||
|
||||
if (asrText.isEmpty()) {
|
||||
resetToWaitSpeech()
|
||||
Log.d(TAG, "识别不出来asr")
|
||||
return
|
||||
} else {
|
||||
Log.d(TAG, "识别出来:${asrText}")
|
||||
}
|
||||
|
||||
|
||||
// 评分逻辑
|
||||
var score = 0
|
||||
when {
|
||||
@ -276,7 +202,7 @@ class VoiceController(
|
||||
|
||||
audioBuffer.clear()
|
||||
state = VoiceState.UPLOADING
|
||||
onFinalAudio(audio, asrText) // 传递音频和识别文本
|
||||
onFinalAudio(audio) // 传递音频和识别文本
|
||||
}
|
||||
|
||||
/* ================= 播放回调 ================= */
|
||||
@ -330,7 +256,6 @@ class VoiceController(
|
||||
fun release() {
|
||||
wakeupManager.release()
|
||||
vadManager.reset()
|
||||
offlineRecognizer?.release()
|
||||
}
|
||||
|
||||
/* ================= Utils ================= */
|
||||
|
||||
@ -10,6 +10,7 @@ import android.media.AudioFormat
|
||||
import android.media.AudioRecord
|
||||
import android.media.MediaRecorder
|
||||
import android.media.audiofx.AcousticEchoCanceler
|
||||
import android.media.audiofx.NoiseSuppressor
|
||||
import android.os.Bundle
|
||||
import android.os.Environment
|
||||
import android.os.IBinder
|
||||
@ -88,7 +89,7 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
private var voiceController: VoiceController? = null
|
||||
private var audioRecord: AudioRecord? = null
|
||||
private var isRecording = false
|
||||
private val audioSource = MediaRecorder.AudioSource.VOICE_COMMUNICATION
|
||||
private val audioSource = MediaRecorder.AudioSource.MIC
|
||||
private val sampleRateInHz = 16000
|
||||
private val channelConfig = AudioFormat.CHANNEL_IN_MONO
|
||||
private val audioFormat = AudioFormat.ENCODING_PCM_16BIT
|
||||
@ -204,12 +205,12 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
}
|
||||
)
|
||||
},
|
||||
onFinalAudio = { audio, asrText ->
|
||||
Log.d("lrsxx", "检测到语音，长度=${audio.size},文本:${asrText}")
|
||||
mViewModel?.uploadVoice(
|
||||
AudioPcmUtil.pcm16ToBase64(AudioPcmUtil.floatToPcm16(audio)),
|
||||
1
|
||||
)
|
||||
onFinalAudio = { audio ->
|
||||
Log.d("lrsxx", "检测到语音，长度=${audio.size}")
|
||||
// mViewModel?.uploadVoice(
|
||||
// AudioPcmUtil.pcm16ToBase64(AudioPcmUtil.floatToPcm16(audio)),
|
||||
// 1
|
||||
// )
|
||||
loadLocalJsonAndPlay()
|
||||
val file = File(
|
||||
getExternalFilesDir(Environment.DIRECTORY_DOWNLOADS)!!.getAbsolutePath(),
|
||||
@ -267,7 +268,7 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
lifecycleScope.launch(Dispatchers.IO) {
|
||||
// UnityPlayerHolder.getInstance()
|
||||
// .startTalking(msg.content)
|
||||
loadLocalJsonAndPlay()
|
||||
// loadLocalJsonAndPlay()
|
||||
}
|
||||
}
|
||||
}
|
||||
@ -326,6 +327,7 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
return false
|
||||
}
|
||||
enableSystemAec(audioRecord!!)
|
||||
enableSystemNs(audioRecord!!)
|
||||
return true
|
||||
}
|
||||
|
||||
@ -343,8 +345,23 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
|
||||
Log.d("VoiceService", "✅ System AEC enabled")
|
||||
}
|
||||
private var noiseSuppressor: NoiseSuppressor? = null
|
||||
|
||||
|
||||
private fun enableSystemNs(record: AudioRecord) {
|
||||
if (!NoiseSuppressor.isAvailable()) {
|
||||
Log.w("VoiceService", "System NS not available")
|
||||
return
|
||||
}
|
||||
|
||||
noiseSuppressor = NoiseSuppressor.create(record.audioSessionId)
|
||||
noiseSuppressor?.enabled = true
|
||||
|
||||
Log.d("VoiceService", "✅ System NoiseSuppressor enabled")
|
||||
}
|
||||
|
||||
|
||||
// private val audioBuffer = mutableListOf<Float>()
|
||||
//开始录音
|
||||
private fun startRecording() {
|
||||
if (isRecording) return
|
||||
@ -366,7 +383,9 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
if (read > 0) {
|
||||
// 避免每次都 new FloatArray，复用
|
||||
for (i in 0 until read) floatBuf[i] = buf[i] / 32768f
|
||||
voiceController?.acceptAudio(floatBuf.copyOf(read))
|
||||
val gg = floatBuf.copyOf(read)
|
||||
voiceController?.acceptAudio(gg)
|
||||
// audioBuffer.addAll(gg.asList())
|
||||
}
|
||||
}
|
||||
}
|
||||
@ -393,6 +412,13 @@ class MainActivity : BaseViewModelActivity<ActivityMainBinding, MainViewModel>()
|
||||
release()
|
||||
audioRecord = null
|
||||
}
|
||||
// val file = File(
|
||||
// getExternalFilesDir(Environment.DIRECTORY_DOWNLOADS)!!.getAbsolutePath(),
|
||||
// "xxx.wav"
|
||||
// )
|
||||
// AudioDebugUtil.saveFloatPcmAsWav(audioBuffer.toFloatArray(), file)
|
||||
// LogUtils.dTag("audioxx", "WAV saved: ${file.path}")
|
||||
// audioBuffer.clear()
|
||||
}
|
||||
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user