iFLYTEK-MSC-Python-SDK

Introduction

• 中文 | English

• A third-party Python SDK for a iFLYTEK MSC. Using for ASR, TSS, KWS.

Quick Start

• Install MSC Python SDK

  $ pip install git+https://github.com/jm12138/iFLYTEK-MSC-Python-SDK

• Download MSC SDK

  • SDK Download

• Voice Wakeup (KWS) using Mic Input

  import msc
  import pyaudio
  from ctypes import string_at, c_void_p
  from threading import Event
  
  # Audio Stream
  p = pyaudio.PyAudio()
  input_stream = p.open(
      format=pyaudio.paInt16,
      channels=1,
      rate=16000,
      input=True,
      frames_per_buffer=2048
  )
  
  # Set APP ID
  appid = '' 
  
  # Set MSC SDK DLL/SO File Path
  # X86: SDK_DIR/bin/msc.dll
  # X64: SDK_DIR/bin/msc_x64.dll
  sdk_path = ''
  
  # Set MSC Client
  client = msc.MSC(sdk_path=sdk_path, params=f"appid={appid}".encode('UTF-8'))
  
  # Stop Event
  stop_event = Event()
  
  # KSW Callback Function
  def message_callback(
          sessionID: bytes,
          msg: int,
          param1: int,
          param2: int,
          info: c_void_p,
          userData: c_void_p
  ):
      # Print Result
      print('sessionID: ', sessionID.decode('UTF-8'))
      print('msg: ', msg)
      print('param1: ', param1)
      print('param2: ', param2)
      print('info: ', string_at(info, param2).decode('UTF-8'))
      print('userData: ', userData)
  
      # Stop
      stop_event.set()
  
      return 0
  
  # Set IVW Res Path
  # SDK_DIR/bin/msc/res/ivw/wakeupresource.jet
  ivw_res_path = ''
  
  # Start KWS
  client.kws(
      params=f'ivw_res_path=fo|{ivw_res_path}'.encode('UTF-8'),
      message_callback=message_callback,
      stream=input_stream,
      chunk_size=2048,
      user_data=None,
      stop_event=stop_event
  )

    pIvwParam = {
        "rlt": [
            {
                "sid": "undefine",
                "istart": 33,
                "iresid": 119,
                "iresIndex": 2,
                "iduration": 189,
                "nfillerscore": 0,
                "nkeywordscore": 142754,
                "ncm": 1525,
                "keyword": "ding1 dong1 ding1 dong1  "
            }
        ]
    }
    data->rlt = {
        "rlt": [
            {
                "sid": "undefine",
                "istart": 33,
                "iresid": 119,
                "iresIndex": 2,
                "iduration": 189,
                "nfillerscore": 0,
                "nkeywordscore": 142754,
                "ncm": 1525,
                "keyword": "ding1 dong1 ding1 dong1  "
            }
        ]
    }
    data->datalen=76864
    sessionID:  civw1PdHHBHvGQRgCahtaaBJvtTw1WLl5L
    msg: 1
    param1: 0
    param2: 98
    info: {
        "sst": "wakeup",
        "id": 2,
        "score": 1525,
        "bos": 330,
        "eos": 2220,
        "keyword": "lao3-zhang1-lao3-zhang1"
    }
    userData:  None
    wakeup : ivw param = {
        "rlt": [
            {
                "sid": "undefine",
                "istart": 33,
                "iresid": 119,
                "iresIndex": 2,
                "iduration": 189,
                "nfillerscore": 0,
                "nkeywordscore": 142754,
                "ncm": 1525,
                "keyword": "ding1 dong1 ding1 dong1  "
            }
        ]
    }
  

• Speech Recognizer (ASR) using Mic Input

  import msc
  import pyaudio
  
  # Audio Stream
  p = pyaudio.PyAudio()
  input_stream = p.open(
      format=pyaudio.paInt16, channels=1, rate=16000, input=True, frames_per_buffer=2048
  )
  
  # Set APP ID
  appid = '' 
  
  # Set MSC SDK DLL/SO File Path
  # X86: SDK_DIR/bin/msc.dll
  # X64: SDK_DIR/bin/msc_x64.dll
  sdk_path = ''
  
  # Set MSC Client
  client = msc.MSC(sdk_path=sdk_path, params=f"appid={appid}".encode('UTF-8'))
  
  # Set Domain
  domain = "iat"
  
  # Start ASR
  for item in client.asr(
      params=f"domain={domain}".encode("UTF-8"),
      stream=input_stream,
      chunk_size=2048,
  ):
      print(item.decode("UTF-8"))

    {"sn":1,"ls":false,"bg":0,"ed":0,"ws":[{"bg":64,"cw":[{"sc":0.0,"w":"今天"}]},{"bg":132,"cw":[{"sc":0.0,"w":"天气"}]},{"bg":164,"cw":[{"sc":0.0,"w":"怎么样"}]}]}
    {"sn":2,"ls":true,"bg":0,"ed":0,"ws":[{"bg":235,"cw":[{"sc":0.0,"w":"？"}]}]}
  

• Speech Synthesizer (TTS)

  import msc
  import pyaudio
  
  # Audio Stream
  p = pyaudio.PyAudio()
  output_stream = p.open(format=pyaudio.paInt16, channels=1, rate=16000, output=True)
  
  # Set APP ID
  appid = '' 
  
  # Set MSC SDK DLL/SO File Path
  # X86: SDK_DIR/bin/msc.dll
  # X64: SDK_DIR/bin/msc_x64.dll
  sdk_path = ''
  
  # Set MSC Client
  client = msc.MSC(sdk_path=sdk_path, params=f"appid={appid}".encode('UTF-8'))
  
  # Set Text Encoding
  text_encoding = "UTF8"
  
  # Start TTS
  for item in client.tts(
      params=f"text_encoding={text_encoding}".encode("UTF-8"),
      text="你好，我是您的语音助手，有什么需要帮助的吗？".encode("UTF-8"),
  ):
      output_stream.write(item)

Documentation

• Offical Documents