这是我在这里发布的第二个问题，所以如果我做错了什么，请告诉我。
我今天面临一个有趣的问题。我为呼叫中心工作，我公司的一位客户验证信息，希望从客户那里收集银行帐号，并希望我们的客户服务代理将所述银行帐号输入客户外部网站。
这些银行帐号不会保存在我们本地数据库的任何地方，但我们的 CSR 收集银行帐号的音频将保存在我们的系统中。纯文本将不可用，但声音文件将可用。我的问题是，是否有一种方法可以使用程序自动对录音的某个部分进行自动加扰？我知道这是在黑暗中的严重镜头。谢谢你。
问问题
1130 次
1 回答
3
虽然您的问题不涉及特定的编程相关问题，但我会尝试回答它，因为我正在做类似的事情。
我们可以使用程序自动对录音的某个部分进行自动加扰吗？
我们当然可以。这将取决于您想要制作的复杂程度。
虽然有复杂的方法，但从一个非常基本的概念角度来看，我们将需要在以下阶段对录制的音频文件进行处理。
在音频文件中拆分单词：这将需要单词之间的静音识别。- 通过语音识别系统传递每个单词
- 想出一个打法的办法。是否要静音，
用或
jumble填充它。white noiseencode - 将识别的单词与您想要的单词进行比较
scramble，如果存在scramble基于所选方法 的匹配项 concatenate将所有单词按正确顺序组合 ( ) 并存储。
我已经整理了一个基本原型，除了（4）之外，它可以在上面做。该程序大量使用pydub，它提供了更简单的音频操作方法。可以在此处找到有关该内容的教程。
该程序基本上，
1) 我从这个站点下载了开源预录的 wav 文件0以获取数字10并使用pydub.
该程序以一秒钟的块切割给定的音频文件。我使用了大胆来分隔每个单词，以便它们适合一秒钟的窗口。在现实生活中，情况并非如此。
2）然后将单词通过谷歌语音识别引擎并显示识别的单词。如您所见，该词six无法正确识别。speech recognition engine为此，您将需要一个健壮的。
3) 该程序提供三种不同的scramble方法。
- a) 反转单词
- b）用等价替换单词
white noise - c) 将单词替换为
silence
4）然后选择三个单词9，4并2应用上述scramble方法替换相应的单词文件
5）然后它以正确的顺序连接所有单词以及打乱的单词并创建输出文件。
注意：我没有足够的时间来比较 word to scramble 和识别的 word。
如果有任何问题，请告诉我。
****演示代码：****
""" Declarations """
import speech_recognition as sr
from pydub import AudioSegment
from pydub.silence import split_on_silence
from pydub.generators import WhiteNoise
from pydub.playback import play
""" Function for Speech Recognition """
def processAudio(WAV_FILE):
r = sr.Recognizer()
with sr.WavFile(WAV_FILE) as source:
audio = r.record(source) # read the entire WAV file
# recognize speech using Google Speech Recognition
try:
print("recognizedWord=" + r.recognize_google(audio))
except sr.UnknownValueError:
print("Could not understand audio")
except sr.RequestError as e:
print("Could not request results from GSR; {0}".format(e))
""" Function to scramble word based upon choice """
def scramble_audio(aWord, option):
scramble_file = export_path + "slice" + str(aWord) +".wav"
scramble_audioseg = AudioSegment.from_wav(scramble_file)
aWord_length = scramble_audioseg.__len__() #Get length of word segment to scramble
if option == "reverse": #Reverse word to scramble
scrambled_word = scramble_audioseg.reverse()
elif option == "whiteNoise": #Replace word to scramble with white noise
wn = WhiteNoise() #Instantiate White Noise Object
aWord_length = scramble_audioseg.__len__() #Get length of word segment
scrambled_word = wn.to_audio_segment(duration=aWord_length) #Create audio_segment
elif option == "silence": #Replace word to scramble with silence
scrambled_word = AudioSegment.silent(duration=aWord_length)
print ("Scrambling and Exporting %s" % scramble_file)
scrambled_word.export(scramble_file, format="wav") #Export merged audio file
if __name__ == "__main__":
export_path = ".//splitAudio//"
in_audio_file = "0-10.wav"
out_audio_file = export_path + "scrambledAudio.wav"
#Read main audio file to be processed. Assuming in the same folder as this script
sound = AudioSegment.from_wav(in_audio_file)
sec2_splice = 1 #Splice threshold in sec
audio_length = len(sound) # Total Audio Length In millisec
q, r = divmod(audio_length, sec2_splice) #Get quotient and remainder
#Get total segments and rounds to next greater integer
total_segments= (q + int(bool(r)) ) / 1000 #Converting to sec
#Iterate through slices every one second and export
print ("")
n=0
while n <= total_segments:
print ("Making slice from %d to %d (sec)" % (n , sec2_splice ))
temp_object = sound[ (n * 1000) : (sec2_splice * 1000)] #Slicing is done in millisec
myaudio_file = export_path + "slice" + str(n) +".wav"
temp_object.export(myaudio_file , format="wav")
print ("Trying to recognize %d " %n)
processAudio(myaudio_file)
n = sec2_splice
sec2_splice += 1
#Scramble desired audio slice
print ("")
scramble_word = 9
scramble_audio(scramble_word, "reverse" )
scramble_word = 4
scramble_audio(scramble_word, "whiteNoise" )
scramble_word = 2
scramble_audio(scramble_word, "silence" )
#Combine modified audio
final_audio = AudioSegment.empty() #Create empty AudioSegment
print ("")
i = 0
while i <= total_segments:
temp_audio_file = export_path + "slice" + str(i) +".wav"
temp_audio_seg = AudioSegment.from_wav(temp_audio_file)
print ("Combining %s" % temp_audio_file )
final_audio = final_audio.append(temp_audio_seg, crossfade=0)
i += 1
print ("Exporting final audio %s" % out_audio_file )
final_audio.export(out_audio_file , format="wav")
输出：
Python 2.7.9 (default, Dec 10 2014, 12:24:55) [MSC v.1500 32 bit (Intel)] on win32
Type "copyright", "credits" or "license()" for more information.
>>> ================================ RESTART ================================
>>>
Making slice from 0 to 1 (sec)
Trying to recognize 0
recognizedWord=0
Making slice from 1 to 2 (sec)
Trying to recognize 1
recognizedWord=1
Making slice from 2 to 3 (sec)
Trying to recognize 2
Could not understand audio
Making slice from 3 to 4 (sec)
Trying to recognize 3
recognizedWord=3
Making slice from 4 to 5 (sec)
Trying to recognize 4
recognizedWord=4
Making slice from 5 to 6 (sec)
Trying to recognize 5
recognizedWord=5
Making slice from 6 to 7 (sec)
Trying to recognize 6
recognizedWord=sex
Making slice from 7 to 8 (sec)
Trying to recognize 7
recognizedWord=7
Making slice from 8 to 9 (sec)
Trying to recognize 8
recognizedWord=8
Making slice from 9 to 10 (sec)
Trying to recognize 9
recognizedWord=9
Making slice from 10 to 11 (sec)
Trying to recognize 10
recognizedWord=10
Scrambling and Exporting .//splitAudio//slice9.wav
Scrambling and Exporting .//splitAudio//slice4.wav
Scrambling and Exporting .//splitAudio//slice2.wav
Combining .//splitAudio//slice0.wav
Combining .//splitAudio//slice1.wav
Combining .//splitAudio//slice2.wav
Combining .//splitAudio//slice3.wav
Combining .//splitAudio//slice4.wav
Combining .//splitAudio//slice5.wav
Combining .//splitAudio//slice6.wav
Combining .//splitAudio//slice7.wav
Combining .//splitAudio//slice8.wav
Combining .//splitAudio//slice9.wav
Combining .//splitAudio//slice10.wav
Exporting final audio .//splitAudio//scrambledAudio.wav
>>>
于 2016-04-08T17:36:54.197 回答