IBM Speech-To-Text 的输出

Question

import pyaudio
from ibm_watson import SpeechToTextV1
from ibm_watson.websocket import RecognizeCallback, AudioSource
from threading import Thread
from ibm_cloud_sdk_core.authenticators import IAMAuthenticator

try:
    from Queue import Queue, Full
except ImportError:
    from queue import Queue, Full

###############################################
#### Initalize queue to store the recordings ##
###############################################
CHUNK = 1024
# Note: It will discard if the websocket client can't consumme fast enough
# So, increase the max size as per your choice
BUF_MAX_SIZE = CHUNK * 10
# Buffer to store audio
q = Queue(maxsize=int(round(BUF_MAX_SIZE / CHUNK)))

# Create an instance of AudioSource
audio_source = AudioSource(q, True, True)

###############################################
#### Prepare Speech to Text Service ########
###############################################

# initialize speech to text service
authenticator = IAMAuthenticator('apikey')
speech_to_text = SpeechToTextV1(authenticator=authenticator)
#speech_to_text.set_service_url('https://api.us-south.speech-to-text.watson.cloud.ibm.com/instances/62a2f19f-959f-4c3c-a276-27ab0e458341/v1/recognize')
speech_to_text.set_service_url('https://stream.watsonplatform.net/speech-to-text/api')

# define callback for the speech to text service
class MyRecognizeCallback(RecognizeCallback):
    def __init__(self):
        RecognizeCallback.__init__(self)

    def on_transcription(self, transcript):
        print(transcript)

    def on_connected(self):
        print('Connection was successful')

    def on_error(self, error):
        print('Error received: {}'.format(error))

    def on_inactivity_timeout(self, error):
        print('Inactivity timeout: {}'.format(error))

    def on_listening(self):
        print('Service is listening')

    def on_hypothesis(self, hypothesis):
        print(hypothesis)

    def on_data(self, data):
        print(data)

    def on_close(self):
        print("Connection closed")

# this function will initiate the recognize service and pass in the AudioSource
def recognize_using_weboscket(*args):
    mycallback = MyRecognizeCallback()
    speech_to_text.recognize_using_websocket(audio=audio_source,
                                             content_type='audio/l16; rate=44100',
                                             recognize_callback=mycallback,
                                             interim_results=True)

###############################################
#### Prepare the for recording using Pyaudio ##
###############################################

# Variables for recording the speech
FORMAT = pyaudio.paInt16
CHANNELS = 1
RATE = 44100

# define callback for pyaudio to store the recording in queue
def pyaudio_callback(in_data, frame_count, time_info, status):
    try:
        q.put(in_data)
    except Full:
        pass # discard
    return (None, pyaudio.paContinue)

# instantiate pyaudio
audio = pyaudio.PyAudio()

# open stream using callback
stream = audio.open(
    format=FORMAT,
    channels=CHANNELS,
    rate=RATE,
    input=True,
    frames_per_buffer=CHUNK,
    stream_callback=pyaudio_callback,
    start=False
)

#########################################################################
#### Start the recording and start service to recognize the stream ######
#########################################################################

print("Enter CTRL+C to end recording...")
stream.start_stream()

try:
    recognize_thread = Thread(target=recognize_using_weboscket, args=())
    recognize_thread.start()

    while True:
        pass
except KeyboardInterrupt:
    # stop recording
    stream.stop_stream()
    stream.close()
    audio.terminate()
    audio_source.completed_recording()

这是使用麦克风作为输入的 IBM Speech-To-Text 服务的代码。我可以知道程序的输出是什么吗？这是我得到的输出：

Enter CTRL+C to end recording...
Connection was successful
Service is listening
  File "C:\Users\---\AppData\Local\Programs\Python\Python38-32\lib\site-packages\websocket\_app.py", line 320, in _callback
    callback(self, *args)
  File "C:\Users\---\AppData\Local\Programs\Python\Python38-32\lib\site-packages\ibm_watson\websocket\recognize_listener.py", line 199, in on_data
    hypothesis = json_object['results'][0]['alternatives'][0][
Connection closed

Answer 1

当我用我的无线耳机麦克风进行测试时，它突然起作用了。不知道为什么，因为两个设备都运行良好。输出是控制台中的脚本。

Answer 2

这也发生在我身上，我认为您的问题的原因是您发送到 websocket 的音频可能难以识别，因此 websocket 的响应为无/空，并且当假设函数试图获取回答这个错误是因为结果不存在而发生的。

hypotesis 函数 (def hypotesis) 的输出将是一个带有转录音频文件结果的字符串，而数据函数 (def data) 的输出将是这样的 json：

{'result_index': 0, 'results': [{'final': True, 'alternatives': [{'transcript': 'hello ', 'confidence': 0.66}], 'keywords_result': {}}]}