[問題]inaSpeechSegmenter audio segmentation

[yshihyu (yshihyu)]

https://github.com/ina-foss/inaSpeechSegmenter

我使用這套件去分割 mp3 單字檔案

因為我mp3檔案很多我是掃所有資料夾底下mp3檔案用 inaSpeechSegmenter 分析

但我發現單音mp3 分析時候會出現錯誤, 下面檔案是測試出現錯誤的單音mp3檔案

http://www.mediafire.com/file/krfuk8wshq5jr2o/0021.mp3/file

下面是執行指令

ina_speech_segmenter.py -i 0021.mp3 -o out/

錯誤訊息我看不出來原因～

ValueError: `window_shape` is too large 有關系

下面是完整訊息

Traceback (most recent call last):
  File "/home/shihyu/.local/bin/ina_speech_segmenter.py", line 63, in
<module>
    seg2csv(seg(e), '%s/%s.csv' % (odir, base))
  File
"/home/shihyu/.local/lib/python3.6/site-packages/inaSpeechSegmenter/segmenter.py",
line 174, in __call__
    return self.segmentwav(tmpwav)
  File
"/home/shihyu/.local/lib/python3.6/site-packages/inaSpeechSegmenter/segmenter.py",
line 148, in segmentwav
    data21, finite = _get_patches(mspec[:, :21], 68, 2)
  File
"/home/shihyu/.local/lib/python3.6/site-packages/inaSpeechSegmenter/segmenter.py",
line 69, in _get_patches
    data = vaw(mspec, (w,h), step=step)
  File
"/home/shihyu/.local/lib/python3.6/site-packages/skimage/util/shape.py", line
230, in view_as_windows
    raise ValueError("`window_shape` is too large")
ValueError: `window_shape` is too large

謝謝

--
※ 發信站: 批踢踢實業坊(ptt.cc), 來自: 123.204.6.231 (臺灣)
※ 文章網址: https://www.ptt.cc/bbs/Python/M.1571059820.A.1F6.html
推 sma1033: 應該是因為音檔長度太短吧，你要不要檢查一下失敗的音檔
→ sma1033: 看是不是長度都特別短