我試圖建立一個(gè)音頻可見(jiàn)波形轉(zhuǎn)換器一拉通感，但我發(fā)現(xiàn)了一個(gè)可愛(ài)的小錯(cuò)誤：/Users/nathanielastudillo/anaconda/lib/python3.6/site-packages/scipy/io/wavfile.py:172: WavFileWarning: Chunk (non-data) not understood, skipping it.  WavFileWarning)Traceback (most recent call last):  File "<ipython-input-8-c1948c42b069>", line 5, in <module>    rate, audio = wavfile.read('3 - Headache.wav')  File "/Users/nathanielastudillo/anaconda/lib/python3.6/site-packages/scipy/io/wavfile.py", line 166, in read    data = _read_data_chunk(fid, comp, noc, bits, mmap=mmap)  File "/Users/nathanielastudillo/anaconda/lib/python3.6/site-packages/scipy/io/wavfile.py", line 71, in _read_data_chunk    data = numpy.fromstring(fid.read(size), dtype=dtype)TypeError: data type "<i3" not understood真令人愉快。我對(duì) numpy 的 dtype文檔進(jìn)行了一些挖掘，發(fā)現(xiàn) wavfile.py 難以咀嚼的非數(shù)據(jù)塊是一個(gè)小端有符號(hào)整數(shù)，大概有 3 個(gè)字節(jié)長(zhǎng)，根據(jù)@mgilson 對(duì)這個(gè)問(wèn)題的評(píng)論。無(wú)論出于何種原因，numpy 似乎不想使用 3 字節(jié)小端有符號(hào)整數(shù)。我想使用這里提到的修復(fù)程序，但我什至無(wú)法加載文件。嘗試將修改版本的 wavefile.py 拼湊在一起是否有意義？有誰(shuí)知道我如何解決這個(gè)問(wèn)題？我的實(shí)現(xiàn)，借用this，是：from scipy import signalfrom scipy.io import wavfileimport scipy.io.wavfileM=1024rate, audio = wavfile.read('3 - Headache.wav')freqs, times, spect = signal.spectrogram(audio, fs=rate, window='hanning',                                  nperseg=1024, noverlap=M - 100,                                  detrend=False, scaling='spectrum')