xfei_speech_recog.h

Go to the documentation of this file.

#ifndef XFEI_SPEECH_RECOG_H_
#define XFEI_SPEECH_RECOG_H_
#include <iostream>
#include "file_operation.h"
#include "linuxrec.h"

/* 科大讯飞识别状态和参数 */
struct speech_recog
{
  int ep_stat;
  int rec_stat;     /* 关键词识别状态 */
  int audio_status; /* 录音buffer的状态 */
  volatile int state;
  char* session_begin_params; /* asr各种参数 */
};

/* 科大讯飞语音识别模块 */
class XfeiSpeechRecog
{
public:
  XfeiSpeechRecog() : recog_result_vector(2)
  {
  }

  void setAsrParams(const std::string base_path, const std::string pcm_file, const std::string params,
                    const int channel);

  void initAsr();

  void getPcmFileData();

  void getPCMData(struct DataBuff pcm_buff);

  struct DataBuff recordThroughMIC(const float record_time, bool enable_audio_save);

  void writeAudioData(const char* audio_data, unsigned int audio_len);

  char* dataLoopRecog();

  void uninitAsr();

  void stopRecordThroughMIC();

  std::vector<std::string> resultFromJson();

private:
  char* getRecogResultLoop();

  struct DataBuff getOneChannelData();

  void saveRecordDataToFile();

  int record_loops_ = 0;
  std::vector<std::string> recog_result_vector;
  const char* session_id_ = NULL;
  int audio_stat_;
  struct DataBuff pcm_data_ = { NULL, 0 };  // 存储pcm音频数据
  struct speech_recog speech_recog_;
  char* rec_result_ = NULL;  // 保存语音识别结果
  std::string final_recog_result_;
  int recog_confidence_ = 0;
  std::string base_path_;
  std::string pcm_file_;
  std::string asr_params_;
  int channel_;
  struct recorder* asr_record_ = NULL;
  FileOperation pcm_file_operation;
  RecordAlsaAPI record_alsa;
};

#endif  // XFEI_SPEECH_RECOG_H_