STT: recognition connection failed: [Errno 32] Broken pipe

[GrahamBerends]

Hi All.

I keep getting these broken pipe messages. Some STT conversions work, others don't. So the code basically works.

I imagine that there's a stretch of speech which the speech_recognition doesn't like, so it dumps the whole process with a Broken Pipe.

Can you please help me find some ideas for a work-around:
- How can I identify these parts which the speech_recognition doesn't like? And, say, print "xxxx" wherever there is a section which cannot be converted to text? Or,
- Can you explain how I can maybe "clean" the speech-input-file?

If you need my input audio_file, I will gladly post it.

Look forward to your replies.
Graham

This command works, generating good output:

rm transcript.txt; python ${stt}/WAV_convert_2_text.py ${stt}/Joanne_davilla_35seconds.wav >> transcript.txt

This command fails:

rm transcript.txt; python ${stt}/WAV_convert_2_text.py transcript.wav >> transcript.txt

This is the error output:

transcript.wav
Could not request results from google Speech Recognition service; recognition connection failed: [Errno 32] Broken pipe

This is the code:

import speech_recognition as sr
import sys
from pydub import AudioSegment

# read WAV filename from arguments
audio_file = sys.argv[1]

def readAudioFile(audio_file):
    r = sr.Recognizer()
    file = sr.AudioFile(audio_file)
    with file as source:
 
        audio = r.record(source)
        type(audio)
    try:
        message = r.recognize_google(audio)
        print(message)
        print("Data extracted from phone conversation:\n" + message)
    except sr.UnknownValueError:
        print("Google Speech Recognition could not understand audio")
    except sr.RequestError as e:
        print("Could not request results from google Speech Recognition service; {0}".format(e))
    except IOError as e: 
        print("IOError; {0}".format(e))
    
# parser = argparse.ArgumentParser()
# parser.add_argument("--file", "-f", type=str, required=True)
# args = parser.parse_args()
# audio_file = args.file
print(audio_file)
readAudioFile(audio_file)