파이썬에 대한 Watson을 사용하여 텍스트에 연속 실시간 음성

Question

어떻게 작동하는지와 비슷한 Watson 서버의 마이크를 사용하여 실시간으로 텍스트를 가져올 수있는 작은 파이썬 프로그램을 만들려고합니다. here.

이

내가 생각 해낸 한 코드입니다하지만 녹음을 마친 후이 텍스트를 가져옵니다

내가 나서서 WebSocket을을 사용하여 왓슨 서버에 연결하는 데 처음부터 프로그램을 만들어

import pyaudio 
import json 
from watson_developer_cloud import SpeechToTextV1 

CHUNK = 1024 
FORMAT = pyaudio.paInt16 
CHANNELS = 2 
RATE = 44100 
RECORD_SECONDS = 10 

p = pyaudio.PyAudio() 

stream = p.open(format=FORMAT, 
       channels=CHANNELS, 
       rate=RATE, 
       input=True, 
       frames_per_buffer=CHUNK) 
print("* recording") 

frames = [] 

for i in range(0, int(RATE/CHUNK * RECORD_SECONDS)): 
    data = stream.read(CHUNK) 
    frames.append(data) 

print("* done recording") 

stream.stop_stream() 
stream.close() 
p.terminate() 

data_feed = b''.join(frames) 

speech_to_text = SpeechToTextV1(
    username='secret', 
    password='secret too', 
    x_watson_learning_opt_out=False 
) 

result = speech_to_text.recognize(data_feed, 
            content_type="audio/l16;rate=44100;channels=2", 
            word_confidence=True, 
            max_alternatives=4, 
            word_alternatives_threshold=0.5, 
            model="en-US_BroadbandModel", 
            continuous=True) 

j = json.dumps(result, indent=2) 
print(j) 

Answer 1

. 그것은 여전히 ​​내가 기대하는 바를 정확히 수행하지는 못하지만 아주 가깝습니다.

실시간으로 오디오가 서버로 전송되지만 녹음이 끝나면 녹음을 듣고 있습니다.

import asyncio 
import websockets 
import json 
import requests 
import pyaudio 
import time 

# Variables to use for recording audio 
CHUNK = 1024 
FORMAT = pyaudio.paInt16 
CHANNELS = 2 
RATE = 16000 

p = pyaudio.PyAudio() 

# This is the language model to use to transcribe the audio 
model = "en-US_BroadbandModel" 

# These are the urls we will be using to communicate with Watson 
default_url = "https://stream.watsonplatform.net/speech-to-text/api" 
token_url = "https://stream.watsonplatform.net/authorization/api/v1/token?" \ 
      "url=https://stream.watsonplatform.net/speech-to-text/api" 
url = "wss://stream.watsonplatform.net/speech-to-text/api/v1/recognize?model=en-US_BroadbandModel" 

# BlueMix app credentials 
username = "" # Your Bluemix App username 
password = "" # Your Bluemix App password 

# Send a request to get an authorization key 
r = requests.get(token_url, auth=(username, password)) 
auth_token = r.text 
token_header = {"X-Watson-Authorization-Token": auth_token} 

# Params to use for Watson API 
params = { 
    "word_confidence": True, 
    "content_type": "audio/l16;rate=16000;channels=2", 
    "action": "start", 
    "interim_results": True 
} 

# Opens the stream to start recording from the default microphone 
stream = p.open(format=FORMAT, 
       channels=CHANNELS, 
       rate=RATE, 
       input=True, 
       output=True, 
       frames_per_buffer=CHUNK) 


async def send_audio(ws): 
    # Starts recording of microphone 
    print("* READY *") 

    start = time.time() 
    while True: 
     try: 
      print(".") 
      data = stream.read(CHUNK) 
      await ws.send(data) 
      if time.time() - start > 20: # Records for n seconds 
       await ws.send(json.dumps({'action': 'stop'})) 
       return False 
     except Exception as e: 
      print(e) 
      return False 

    # Stop the stream and terminate the recording 
    stream.stop_stream() 
    stream.close() 
    p.terminate() 


async def speech_to_text(): 
    async with websockets.connect(url, extra_headers=token_header) as conn: 
     # Send request to watson and waits for the listening response 
     send = await conn.send(json.dumps(params)) 
     rec = await conn.recv() 
     print(rec) 
     asyncio.ensure_future(send_audio(conn)) 

     # Keeps receiving transcript until we have the final transcript 
     while True: 
      try: 
       rec = await conn.recv() 
       parsed = json.loads(rec) 
       transcript = parsed["results"][0]["alternatives"][0]["transcript"] 
       print(transcript) 
       #print(parsed) 
       if "results" in parsed: 
        if len(parsed["results"]) > 0: 
         if "final" in parsed["results"][0]: 
          if parsed["results"][0]["final"]: 
           #conn.close() 
           #return False 
           pass 
      except KeyError: 
       conn.close() 
       return False 

# Starts the application loop 
loop = asyncio.get_event_loop() 
loop.run_until_complete(speech_to_text()) 
loop.close() 

이제 마이크를 통해 녹음하는 동안 사본을 얻는 것이 전부입니다.