Android外接音频源回声消除失效问题咨询

问题描述 #

尝试为AudioRecord和AudioTrack实现回声消除，避免播放音频反馈至录音中。已启用回声消除并强制路由至外接音频系统，但仅设备内置麦克风+扬声器组合正常，外接USB麦克风/扬声器、车载蓝牙麦克风/扬声器仍存在音频反馈，回声消除未生效。设备已确认支持AcousticEchoCanceler与NoiseSuppressor。

当前实现代码 #

音频路由设置代码 #

val audioManager = getSystemService(Context.AUDIO_SERVICE) as AudioManager
audioManager.mode = AudioManager.MODE_IN_COMMUNICATION
audioManager.isSpeakerphoneOn = false
audioManager.isWiredHeadsetOn = true // 注意：此属性为只读，手动赋值无效

val bluetoothScoConnected = audioManager.isBluetoothScoOn
val wiredHeadsetConnected = audioManager.isWiredHeadsetOn
val isUsbHeadset = true
val earpieceMode = false

val targetTypes = ArrayList<Int>()
//add types according to needs, may be few in order of importance
if (bluetoothScoConnected) {
    targetTypes.add(AudioDeviceInfo.TYPE_BLUETOOTH_SCO)
} else if (wiredHeadsetConnected) {
    if (isUsbHeadset) {
        targetTypes.add(AudioDeviceInfo.TYPE_USB_HEADSET)
        targetTypes.add(AudioDeviceInfo.TYPE_USB_DEVICE)
        targetTypes.add(AudioDeviceInfo.TYPE_USB_ACCESSORY)
    } else {
        targetTypes.add(AudioDeviceInfo.TYPE_WIRED_HEADSET)
        targetTypes.add(AudioDeviceInfo.TYPE_WIRED_HEADPHONES)
    }
} else if (earpieceMode) {
    targetTypes.add(AudioDeviceInfo.TYPE_BUILTIN_EARPIECE)
} else { // play out loud
    targetTypes.add(AudioDeviceInfo.TYPE_BUILTIN_SPEAKER)
}

var result: Boolean? = null
val devices = audioManager.availableCommunicationDevices

Timber.i("Number of devices: ${devices.size}")
outer@ for (targetType in targetTypes) {
    for (device in devices) {
        Timber.i("device: ${device.type}")
        if (device.type == targetType) {
            result = audioManager.setCommunicationDevice(device)
            Timber.i("setCommunicationDevice type:$targetType result:$result")
            break@outer
        }
    }
}

if (result == null) {
    Timber.i("setCommunicationDevice targetType NOT FOUND!!")
}

回声消除初始化代码 #

if (AcousticEchoCanceler.isAvailable())
{
    val echoCanceler = AcousticEchoCanceler.create(audioRecord.audioSessionId);
    echoCanceler.enabled = true
}
if (NoiseSuppressor.isAvailable())
{
    val noiseSuppressor = NoiseSuppressor.create(audioRecord.audioSessionId)
    noiseSuppressor.enabled = true
}

解决方案 #

1. 确保AudioRecord与AudioTrack共享同一音频会话ID #

回声消除需要获取播放端的音频数据来抵消录音中的回声，必须让AudioRecord和AudioTrack使用同一个音频会话ID：

// 创建AudioRecord时获取sessionId
val audioRecord = AudioRecord(MediaRecorder.AudioSource.VOICE_COMMUNICATION, 
                            sampleRate, 
                            channelConfig, 
                            audioFormat, 
                            bufferSize)
val sessionId = audioRecord.audioSessionId

// 创建AudioTrack时传入该sessionId
val audioTrack = AudioTrack(AudioManager.STREAM_VOICE_CALL, 
                          sampleRate, 
                          channelConfig, 
                          audioFormat, 
                          bufferSize, 
                          AudioTrack.MODE_STREAM, 
                          sessionId)

2. 修正音频路由设置中的错误 #

• 移除audioManager.isWiredHeadsetOn = true，该属性是只读状态，手动赋值不会改变实际路由。
• 蓝牙SCO设备需要先启动连接再检测：

// 若要使用蓝牙SCO，先启动连接
if (!audioManager.isBluetoothScoOn) {
    audioManager.startBluetoothSco()
    // 注册广播接收器监听SCO连接状态，连接成功后再执行设备路由设置
}

3. 确认外接设备的回声消除支持能力 #

部分外接USB/蓝牙设备本身不支持硬件回声消除，可通过以下方式检查：

for (device in audioManager.availableCommunicationDevices) {
    val profiles = device.audioProfiles
    for (profile in profiles) {
        if (profile.profileType == AudioProfile.PROFILE_COMMUNICATION) {
            Timber.i("设备${device.type}支持通信模式，大概率支持回声消除")
        }
    }
}

如果设备不支持硬件回声消除，需考虑实现软件回声消除（如使用WebRTC的音频处理库）。

4. 调整回声消除初始化时机 #

必须在AudioRecord成功启动后再初始化AcousticEchoCanceler：

if (audioRecord.state == AudioRecord.STATE_INITIALIZED) {
    audioRecord.startRecording()
    if (AcousticEchoCanceler.isAvailable()) {
        val echoCanceler = AcousticEchoCanceler.create(audioRecord.audioSessionId)
        echoCanceler.enabled = true
    }
}

5. 保证音频参数一致 #

AudioRecord和AudioTrack的采样率、声道数、音频格式必须完全一致，否则回声消除算法无法匹配音频数据：

• 统一使用MediaRecorder.AudioSource.VOICE_COMMUNICATION作为录音源
• 采样率建议使用44100Hz或48000Hz（多数设备支持的标准采样率）
• 声道统一使用单声道（AudioFormat.CHANNEL_IN_MONO/AudioFormat.CHANNEL_OUT_MONO）
• 音频格式统一使用AudioFormat.ENCODING_PCM_16BIT

内容的提问来源于stack exchange，提问作者Simon Rex