TRTC 接入实时语音识别

#import<QCloudSDK/QCloudSDK.h>
//1.使用第三方外部数据源传入语音数据，自定义 data source 需要实现 QCloudAudioDataSource 协议
QDAudioDataSource *dataSource = [[QDAudioDataSource alloc] init];
?
//2.创建 QCloudRealTimeRecognizer 识别实例
QCloudRealTimeRecognizer *realTimeRecognizer = [[QCloudRealTimeRecognizer alloc] initWithConfig:config dataSource:dataSource];

@interface QDAudioDataSource : NSObject<QCloudAudioDataSource>
@end
?
@implementation QDAudioDataSource
@synthesize running = _running;
?
//SDK 会调用此方法获取当前状态
- (BOOL)running{
    return _recording;
}
?
//SDK 会调用 start 方法，实现此协议的类需要初始化数据源
- (void)start:(void(^)(BOOL didStart, NSError *error))completion{
  _data = [[NSMutableData alloc] init];
   completion(YES,nil);
}
?
//SDK 会调用 stop 方法，实现此协议的类需要停止提供数据
- (void)stop{
  _recording = NO;
  _data = nil;
}
?
//SDK 会调用实现此协议的对象的此方法读取语音数据
- (nullable NSData *)readData:(NSInteger)expectLength{
  NSData *data = nil;
  if ([_data length] >= _offset + expectLength) {
      data = [_data subdataWithRange:NSMakeRange(_offset, expectLength)];
      [_data replaceBytesInRange:NSMakeRange(_offset, expectLength) withBytes:NULL length:0];
  }
  return data;
}
?
//此处仅为演示，需用户自行完善音频数据源填充
- (void)didRecordAudioData:(void * const )bytes length:(NSInteger)length{
  [_data appendBytes:bytes length:length];
}
@end

//1.TRTCCloudListener.TRTCAudioFrameListener 是 TRTC 获取本地麦克风采集到的音频数据回调??由于 ASR 识别16k或8k采样率的音频数据，所以需要设置 setAudioQuality 为 TRTCCloudDef#TRTC_AUDIO_QUALITY_SPEECH?（流畅：采样率：16k；单声道；音频裸码率：16kbps）
void?onCapturedRawAudioFrame(TRTCCloudDef.TRTCAudioFrame trtcAudioFrame)?{
dataSource.writeByte(bytesToShort(trtcAudioFrame.data));; 
} 
//以下方法为把 bytes 数组转成 short 数组
public static short[] bytesToShort(byte[] bytes) {
????if (bytes == null) {
????????return null;
????}
????short[] shorts = new short[bytes.length / 2];
ByteBuffer.wrap(bytes).order(ByteOrder.LITTLE_ENDIAN).asShortBuffer().get(shorts);
????return shorts;
}

//1.使用第三方外部数据源传入语音数据，自定义 data source 需要实现 PcmAudioDataSource 接口
AudioDataSource dataSource = new AudioDataSource(); 
//2.初始化语音识别请求
final?AudioRecognizeRequest?audioRecognizeRequest = new AudioRecognizeRequest.Builder()
.pcmAudioDataSource(dataSource)
.build(); 

private ConcurrentLinkedDeque<Short> shortList = new ConcurrentLinkedDeque<>();
private static boolean first;
?
public class AudioDataSource implements PcmAudioDataSource {?
//向语音识别器添加数据，将长度为?length 的数据从下标0开始复制到?audioPcmData 数组中，并返回实际的复制的数据量的长度?
@Override
????public int read(short[] audioPcmData, int length) {
????????short[] tempData = new short[length];
????????try {
????????????if (!first) {
????????????????first = true;
????????????????SystemClock.sleep(1000);
????????????}
????????????ConcurrentLinkedDeque<Short> shorts;
????????????shorts = getDataList();
????????????if (shorts.size() < length) {
????????????????SystemClock.sleep(300);
????????????}
????????????shorts = getDataList();
????????????for (int i = 0; i < tempData.length; i++) {
????????????????tempData[i] = shorts.poll();
????????????}
????????????System.arraycopy(tempData, 0, audioPcmData, 0, tempData.length);
????????} catch (Exception e)?{
????????????return 0;
????????}
????????return tempData.length;
????}?
//启动识别时回调函数，用户可以在这里做些初始化的工作。
@Override
????public void start()?{
????}?
//结束识别时回调函数，用户可以在这里进行一些清理工作
?@Override
????public void stop() {
????}?
//此处仅为演示，需用户自行完善音频数据源填充
public void writeByte(short[] pcmData) {
????????for (short pcmDatum : pcmData) {
????????????shortList.add(pcmDatum);
????????}
????} ?
private ConcurrentLinkedDeque<Short> getDataList() {
????????return shortList;
????}?