WAVファイルの無音部分を区切りとしてファイルを分割する

sample_extractor.py

'''
usage: python3 ./sample_sxtractor *.wav
'''

import sys
import copy
import wave
import numpy as np
from scipy import signal
# import matplotlib.pyplot as plt
# import matplotlib.transforms as mtransforms

LEN_FILT = 10000
THRESH = 200.

if __name__ == "__main__":
    filename = sys.argv[1]
    wave_ary = []
    indata_params = []

    with wave.open(filename, "r") as in_data:
        smp_num_frames = in_data.getnframes()
        indata_params = in_data.getparams()
        wave_ary = in_data.readframes(smp_num_frames)
        wave_ary = np.frombuffer(wave_ary, dtype="int16")
        wave_ary = wave_ary.reshape((-1,2))

        wave_mix = np.sum(wave_ary, axis=1) * 0.5
        wave_abs = np.abs(wave_mix)

        wave_flt = signal.convolve(wave_abs, np.ones(LEN_FILT), mode='same') / LEN_FILT
        wave_thr = np.array(wave_flt >= THRESH, dtype="int16") * LEN_FILT

        wave_flt2 = signal.convolve(wave_thr, np.ones(LEN_FILT), mode='same') / LEN_FILT
        wave_thr2 = np.array(wave_flt2 >= THRESH)

        index = []
        tmp = [0, 0]
        for i in range(len(wave_thr2)-1):
            if (wave_thr2[i] == False) and (wave_thr2[i+1] == True):
                tmp[0] = i
            if (wave_thr2[i] == True) and (wave_thr2[i+1] == False):
                tmp[1] = i
                index.append(copy.copy(tmp))
        
        print(index)

        # fig, ax = plt.subplots()
        # trans = mtransforms.blended_transform_factory(ax.transData, ax.transAxes)

        # ax.plot(wave_abs, )
        # ax.fill_between(np.arange(smp_num_frames), 0, smp_num_frames, wave_thr2,
        #                 facecolor='red', alpha=0.5, transform=trans)
        # ax.set_ylim([0, np.max(wave_abs)*1.2])
        # plt.show()
    count = 0
    filename_split = filename.split(".")
    for i in index:
        filename_split_tmp = copy.copy(filename_split)
        filename_split_tmp.insert(-1, str(count))
        out_data = wave.Wave_write(".".join(filename_split_tmp))
        out_data.setparams(indata_params)
        out_data.writeframes(wave_ary[i[0]:i[1]])
        count += 1