利用headset profile实现语音采集

一、    利用headset profile进行语音采集

在android audio系统中，已经对AUDIO_DEVICE_IN_BLUETOOTH_SCO_HEADSET作了规划，主要是针对headset profile实现语音输入，所以我们需要实现针对该device输入的sco HAL。即要求实现headset profile的mic输入。

1.      修改audio_policy.conf,添加遥控器mic设备以及对应的HAL；HAL命名为audio.sco.vendor.so
#bluetoothmic for headset
  sco{ // HAL的名字
    inputs {
      mvs {
        sampling_rates 16000
        channel_masks AUDIO_CHANNEL_IN_MONO
        formats AUDIO_FORMAT_PCM_16_BIT
        devices AUDIO_DEVICE_IN_BLUETOOTH_SCO_HEADSET
      }
    }
  }
2.      音频策略的修改，由于android 原生音频策略已经规划了AUDIO_DEVICE_IN_BLUETOOTH_SCO_HEADSET，所以不用修改音频策略。

二、headsetprofile建立音频通路的过程

1.     headset profile建立连接的过程

2.     建立语音通路的过程

建立语音通路的前提是headset device已经建立连接，并进入connected状态。AudioManager里建立sco连接的api：

    /**

     *Start bluetooth SCO audio connection.

     *<p>Requires Permission:

    *   {@linkandroid.Manifest.permission#MODIFY_AUDIO_SETTINGS}.

     *<p>This method can be used by applications wanting to send and receivedaudio

     *to/from a bluetooth SCO headset while the phone is not in call.

     *<p>As the SCO connection establishment can take several seconds,

     *applications should not rely on the connection to be available when the method

     *returns but instead register to receive the intent {@link#ACTION_SCO_AUDIO_STATE_UPDATED}

     *and wait for the state to be {@link #SCO_AUDIO_STATE_CONNECTED}.

     *<p>As the ACTION_SCO_AUDIO_STATE_UPDATED intent is sticky, theapplication can check the SCO

     *audio state before calling startBluetoothSco() by reading the intent returnedby the receiver

     *registration. If the state is already CONNECTED, no state change will bereceived via the

     *intent after calling startBluetoothSco(). It is however useful to callstartBluetoothSco()

     *so that the connection stays active in case the current initiator stops theconnection.

     *<p>Unless the connection is already active as described above, the statewill always

     *transition from DISCONNECTED to CONNECTING and then either to CONNECTED if theconnection

     *succeeds or back to DISCONNECTED if the connection fails (e.g no headset isconnected).

     *<p>When finished with the SCO connection or if the establishment fails,the application must

     *call {@link #stopBluetoothSco()} to clear the request and turn down thebluetooth connection.

     *<p>Even if a SCO connection is established, the following restrictionsapply on audio

    * output streams so that they can be routed to SCO headset:

    * <ul>

    *   <li> the stream typemust be {@link #STREAM_VOICE_CALL} </li>

    *   <li> the format must bemono </li>

    *   <li> the sampling mustbe 16kHz or 8kHz </li>

     *</ul>

     * <p>The following restrictions apply on input streams:

    * <ul>

    *   <li> the format must bemono </li>

    *   <li> the sampling mustbe 8kHz </li>

    * </ul>

     *<p>Note that the phone application always has the priority on the usageof the SCO

     *connection for telephony. If this method is called while the phone is in call

     *it will be ignored. Similarly, if a call is received or sent while anapplication

     *is using the SCO connection, the connection will be lost for the applicationand NOT

     *returned automatically when the call ends.

     *<p>NOTE: up to and including API version

     *{@link android.os.Build.VERSION_CODES#JELLY_BEAN_MR1}, this method initiates avirtual

     *voice call to the bluetooth headset.

     *After API version {@link android.os.Build.VERSION_CODES#JELLY_BEAN_MR2} only a raw SCO audio

     *connection is established.

     *@see #stopBluetoothSco()

     *@see #ACTION_SCO_AUDIO_STATE_UPDATED

     */

   public void startBluetoothSco(){

        IAudioService service = getService();

       try {

           service.startBluetoothSco(mICallBack,

                   getContext().getApplicationInfo().targetSdkVersion);

       } catch (RemoteException e) {

           throw e.rethrowFromSystemServer();

       }

    }

3.     使用AudioRecord捕获原始音频流，我们可以对采集的数据进行降噪合成等处理，这刚好符合语音识别的语音采集要求。AudioRecord采集流程很简单，我们只需要构造一个AudioRecord对象，然后传入各种不同配置的参数即可。