我正在制作一个将麦克风录制的音频转换为文本的应用程序。录音的时间可能会很长,例如3个小时,所以我想将其转换为持续时间较短(例如一分钟左右)的波形文件,然后产生一个子线程来执行音频到文本操作会更好,而主线程可以在下一分钟开始记录。音频到文本的操作比录制部分要快得多,因此定时不会成为问题。
这是我认为应如何工作的流程图。
我正在pyaudio
用来录制音频。它的代码是:
import pyaudio
import wave
import time
def read_audio(stream):
chunk = 1024 # Record in chunks of 1024 samples
sample_format = pyaudio.paInt16 # 16 bits per sample
channels = 2
fs = 44100 # Record at 44100 samples per second
seconds = 10
filename = 'record.wav'
frames = [] # Initialize array to store frames
# Store data in chunks for 3 seconds
for i in range(0, int(fs / chunk * seconds)):
data = stream.read(chunk)
frames.append(data)
# Save the recorded data as a WAV file
wf = wave.open(filename, 'wb')
wf.setnchannels(channels)
wf.setsampwidth(p.get_sample_size(sample_format))
wf.setframerate(fs)
wf.writeframes(b''.join(frames))
wf.close()
# Stop and close the stream
stream.stop_stream()
stream.close()
p = pyaudio.PyAudio() # Create an interface to PortAudio
chunk = 1024 # Record in chunks of 1024 samples
sample_format = pyaudio.paInt16 # 16 bits per sample
channels = 2
fs = 44100
stream = p.open(format=sample_format,channels=channels,rate=fs,
frames_per_buffer=chunk,input=True)
read_audio(stream)
p.terminate() # Terminate the PortAudio interface
对于语音识别,使用Google的API speech_recognition
。其代码:
import speech_recognition as sr
def convert():
sound = "record.wav"
r = sr.Recognizer()
with sr.AudioFile(sound) as source:
r.adjust_for_ambient_noise(source)
print("Converting Audio To Text and saving to file..... ")
audio = r.listen(source)
try:
value = r.recognize_google(audio) ##### API call to google for speech recognition
if str is bytes:
result = u"{}".format(value).encode("utf-8")
else:
result = "{}".format(value)
with open("test.txt","a") as f:
f.write(result)
print("Done !\n\n")
except sr.UnknownValueError:
print("")
except sr.RequestError as e:
print("{0}".format(e))
except KeyboardInterrupt:
pass
convert()
由于GIL,Python从来就不是真正的多线程,但是在您的情况下这可能并不重要,因为您正在使用api调用来为您进行语音识别。
因此,您可以尝试启动线程进行转换
from threading import Thread
t = Thread(target=convert)
t.start()
在尝试转换下一分钟之前,您可以尝试加入最后一个线程以确保它已完成
t.join()
您可能还可以使用asyncio库
尽管这可能是过大的,但我可能会使用多处理库。在您的情况下,您可能有一个监听程序工作进程正在不断记录和保存新的声音文件,而一个转换工作程序进程正在不断寻找新文件并进行转换。
如果需要的话,这将允许您编写一个更强大的系统。例如,如果您失去了互联网连接,并且在几分钟之内无法通过Google api转换声音文件,录音机工作人员将继续保存声音文件而无需关心,当恢复互联网连接时,声音文件将得到处理。
无论如何,这是您可以使用的转换工作程序流程的一个小示例。
import multiprocessing as mp
import os
from pathlib import Path
from time import sleep
class ConversionWorker:
def __init__(self, sound_file_directory_path: str, text_save_filepath: str):
self.sound_directory_path = Path(sound_file_directory_path)
self.text_filepath = Path(text_save_filepath)
def run(self):
while True:
# find and convert all wav files in the target directory
filepaths = self.sound_directory_path.glob('*.wav')
for path in filepaths:
# convert from path
# save to self.text_filepath
convert()
# we can delete the sound file after converting it
os.remove(path)
# sleep for a bit since we are only saving files once a minute or so
sleep(5)
def main():
conversion_worker = ConversionWorker(sound_file_directory_path='path/to/sounds', text_save_filepath='path/to/text')
p = mp.Process(target=conversion_worker.run)
p.start()
# do the recording and saving for as long as you want
p.terminate()
本文收集自互联网,转载请注明来源。
如有侵权,请联系 [email protected] 删除。
我来说两句