通过麦克风进行 Azure 语音连续语音识别

问题描述 投票：0回答：1

我目前正在尝试编写一个Python脚本，使用连续识别来识别来自麦克风的语音。我使用了 Azure 语音服务中的示例代码（https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-recognize-speech?pivots=programming-language-蟒蛇）。但是，我的程序永远不会退出 while 循环。如何在不输入命令的情况下停止识别？是否可以通过口头命令（例如长时间停顿或说出关键字）来停止连续语音识别？我正在尝试构建一个语音机器人。我是否正确，为了与语音机器人交互，用户应该说话 < 15 s (using single-shot recognition) or interact with the device after every utterance (using continuous recognition) ? Thank you!

重复问题而不回复： https://learn.microsoft.com/en-us/answers/questions/1850234/stop-continuous-speech-recognition-from-microphone?comment=question#newest-question-comment

代码：

import time
from dotenv import dotenv_values
import azure.cognitiveservices.speech as speechsdk

def recognised_speech(evt):
    print(f"You: {evt.result.text}")

def cont_speech_to_text():
    done_talking=False
    
    def stop_cb(evt):
        print('You: {}'.format(evt))
        nonlocal done_talking
        done_talking = True
        speech_recognizer.stop_continuous_recognition()

    #speech_recognizer.recognizing.connect(lambda evt: print('RECOGNIZING: {}'.format(evt)))
    speech_recognizer.recognized.connect(recognised_speech)
    speech_recognizer.session_started.connect(lambda evt: print('SESSION STARTED: {}'.format(evt)))
    speech_recognizer.session_stopped.connect(lambda evt: print('SESSION STOPPED {}'.format(evt)))
    speech_recognizer.canceled.connect(lambda evt: print('CANCELED {}'.format(evt)))

    speech_recognizer.session_stopped.connect(stop_cb)
    speech_recognizer.canceled.connect(stop_cb)

    speech_recognizer.start_continuous_recognition()
    while not done_talking:
        time.sleep(.5)

SPEECH_REGION = "westeurope"
keypath="..."
speechkey=dotenv_values(keypath+".key")
speech_config = speechsdk.SpeechConfig(subscription=speechkey['KEY'], region=SPEECH_REGION)

speech_config.speech_recognition_language="en-US"
speech_config.speech_synthesis_voice_name='en-US-AvaMultilingualNeural'
audio_config = speechsdk.audio.AudioConfig(use_default_microphone=True)
speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)

cont_speech_to_text()

'''

python-3.x speech-recognition azure-speech
1个回答
0
投票

是的，您可以退出循环或停止语音命令或长时间暂停的识别。

  1. 使用语音命令，配置您的识别功能，如下所示。
def cont_speech_to_text():
    done_talking=False

    def recognised_speech(evt):
        print(f"You: {evt.result.text}")
        if "stop listening" in evt.result.text.lower():
            speech_recognizer.stop_continuous_recognition()
    
    def stop_cb(evt):
        print('CLOSING {}'.format(evt))
        speech_recognizer.stop_continuous_recognition()
        nonlocal done_talking
        done_talking = True


    #speech_recognizer.recognizing.connect(lambda evt: print('RECOGNIZING: {}'.format(evt)))
    speech_recognizer.recognized.connect(recognised_speech)
    speech_recognizer.session_started.connect(lambda evt: print('SESSION STARTED: {}'.format(evt)))
    speech_recognizer.session_stopped.connect(lambda evt: print('SESSION STOPPED {}'.format(evt)))
    speech_recognizer.canceled.connect(lambda evt: print('CANCELED {}'.format(evt)))

    speech_recognizer.session_stopped.connect(stop_cb)
    speech_recognizer.canceled.connect(stop_cb)
    

    speech_recognizer.start_continuous_recognition()
    while not done_talking:
        time.sleep(.5)

SPEECH_REGION = "eastus"
speech_config = speechsdk.SpeechConfig(subscription="xxxxxxx", region=SPEECH_REGION)

speech_config.speech_recognition_language="en-US"
speech_config.speech_synthesis_voice_name='en-US-AvaMultilingualNeural'
audio_config = speechsdk.audio.AudioConfig(use_default_microphone=True)
speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)

输出：

SESSION STARTED: SessionEventArgs(session_id=ee53dddc86fc4142ab6a3dddf03c5b7e)
You: Stop listening.
SESSION STOPPED SessionEventArgs(session_id=ee53dddc86fc4142ab6a3dddf03c5b7e)
CLOSING SessionEventArgs(session_id=ee53dddc86fc4142ab6a3dddf03c5b7e)

在这里，我使用了停止监听作为语音命令来停止连续语音识别，但你可以使用任何你想要的。

  1. 每次输入语音后使用暂停时间。

代码

def cont_speech_to_text():
    done_talking=False
    last_recognized_time = time.time()
    pause_threshold = 10

    def recognised_speech(evt):
        print(f"You: {evt.result.text}")
        nonlocal last_recognized_time
        last_recognized_time = time.time()

        # if "stop listening" in evt.result.text.lower():
        #     speech_recognizer.stop_continuous_recognition()
    
    def stop_cb(evt):
        print('CLOSING {}'.format(evt))
        speech_recognizer.stop_continuous_recognition()
        nonlocal done_talking
        done_talking = True


    #speech_recognizer.recognizing.connect(lambda evt: print('RECOGNIZING: {}'.format(evt)))
    speech_recognizer.recognized.connect(recognised_speech)
    speech_recognizer.session_started.connect(lambda evt: print('SESSION STARTED: {}'.format(evt)))
    speech_recognizer.session_stopped.connect(lambda evt: print('SESSION STOPPED {}'.format(evt)))
    speech_recognizer.canceled.connect(lambda evt: print('CANCELED {}'.format(evt)))

    speech_recognizer.session_stopped.connect(stop_cb)
    speech_recognizer.canceled.connect(stop_cb)
    

    speech_recognizer.start_continuous_recognition()
    while not done_talking:
        if time.time() - last_recognized_time > pause_threshold:
            print("Stopping due to long pause...")
            speech_recognizer.stop_continuous_recognition()
        time.sleep(.5)

SPEECH_REGION = "eastus"
speech_config = speechsdk.SpeechConfig(subscription="xxcxxxx", region=SPEECH_REGION)

speech_config.speech_recognition_language="en-US"
speech_config.speech_synthesis_voice_name='en-US-AvaMultilingualNeural'
audio_config = speechsdk.audio.AudioConfig(use_default_microphone=True)
speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)

输出：

SESSION STARTED: SessionEventArgs(session_id=ef6dc34bd7bd4e0b81ae9a6d9f4363c1)
You: Hi.
Stopping due to long pause...
SESSION STOPPED SessionEventArgs(session_id=ef6dc34bd7bd4e0b81ae9a6d9f4363c1)
CLOSING SessionEventArgs(session_id=ef6dc34bd7bd4e0b81ae9a6d9f4363c1)

在这里，我计算每次演讲的时间，并检查 while 循环中的时间差，暂停阈值为 10 秒。

或者您可以同时使用语音命令和暂停时间来停止连续语音识别，上面给出的代码只需取消注释recognised_speech函数中的if条件即可。

最新问题
© www.soinside.com 2019 - 2024. All rights reserved.