测量音频文件开头的静音长度（wav）

Question

我浏览了所有关于测量沉默的帖子，不幸的是我无法找到解决我的问题的方法。

我有大约 3000 个音频文件。每个文件长 10 秒，是某人大声说出一个词的录音。

我需要知道他们花了多长时间才说出这个词（=说出这个词之前的沉默）。我读到我可以使用公式 audioop.rms(fragment, width) 来做到这一点，但我找不到有关如何使用它的说明。

新错误：

void@control:~/Documents$ python audio.py 
Traceback (most recent call last):
  File "audio.py", line 36, in <module>
    leading_silences = {a: get_silence(a, threshold) for a in audio_files}
  File "audio.py", line 36, in <dictcomp>
    leading_silences = {a: get_silence(a, threshold) for a in audio_files}
  File "audio.py", line 7, in get_silence
    song = AudioSegment.from_wav(audio)
  File "/usr/local/lib/python2.7/dist-packages/pydub/audio_segment.py", line 471, in from_wav
    return cls.from_file(file, 'wav')
  File "/usr/local/lib/python2.7/dist-packages/pydub/audio_segment.py", line 387, in from_file
    file = _fd_or_path_or_tempfile(file, 'rb', tempfile=False)
  File "/usr/local/lib/python2.7/dist-packages/pydub/utils.py", line 59, in _fd_or_path_or_tempfile
    fd = open(fd, mode=mode)
IOError: [Errno 2] No such file or directory: 'silbato.wav'

Answer 1

您可以使用 @jiarro 提供的可爱的pydub库：

from pydub import AudioSegment
from os import listdir
from os.path import isfile, join

def get_silence(audio, threshold, interval):
    "get length of silence in seconds from a wav file"

    # swap out pydub import for other types of audio
    song = AudioSegment.from_wav(audio)

    # break into chunks
    chunks = [song[i:i+interval] for i in range(0, len(song), interval)]

    # find number of chunks with dBFS below threshold
    silent_blocks = 0
    for c in chunks:
        if c.dBFS == float('-inf') or c.dBFS < threshold:
            silent_blocks += 1
        else:
            break

    # convert blocks into seconds
    return round(silent_blocks * (interval/1000), 3)

# get files in a directory
audio_path = 'path/to/directory'
audio_files = [i for i in listdir(audio_path) if isfile(join(audio_path, i))]

threshold = -80 # tweak based on signal-to-noise ratio

interval = 1 # ms, increase to speed up

leading_silences = {a: get_silence(join(audio_path, a),
                                   threshold, interval) for a in audio_files}

# to get tab-separated values:
for name, leading_silence in leading_silences.items():
    print(''.join([name, '\t', str(leading_silence)]))