func

import os
import librosa
import numpy as np


# 传入*PHN路径，读取并逐行存在list中
def read_phn(filename):
    f = open(filename, 'r')
    res = list()
    for line in f:
        # line = f.readline()  多余狗屁东西，害我跳行读
        res.append(line.split())
    f.close()
    return res


# 传入DR*的子目录路径, 获取其下的*.PHN, *.WAV的路径
def get_path(path):
    name_list = list()
    for i in os.listdir(path):
        if i[-4:] == '.PHN':
            name_list.append(path + '\\' + i[:-4])
    return name_list


# 传入DR*的子目录路径,将其下WAV按照PHN的帧划分语素，做MFCC处理并保存
def deal_speaker(path, count_dict):
    name_list = get_path(path)
    for name in name_list:
        phn = name + '.PHN'
        wav = name + '.WAV'
        rp = read_phn(phn)
        y, sr = librosa.load(wav, sr=None)
        for i in rp:
            if i[2] in count_dict:
                count_dict[i[2]] += 1
            else:
                count_dict[i[2]] = 1
            if os.path.exists('D:\\语音识别\\LIB\\test_res\\'+i[2]+'\\'):
                np.save('D:\\语音识别\\LIB\\test_res\\'+i[2]+'\\'+str(count_dict[i[2]]), librosa.feature.mfcc(y=y[int(i[0]):int(i[1])], sr=sr, n_mfcc=40))
            else:
                os.mkdir('D:\\语音识别\\LIB\\test_res\\'+i[2]+'\\')
                np.save('D:\\语音识别\\LIB\\test_res\\'+i[2]+'\\'+str(count_dict[i[2]]), librosa.feature.mfcc(y=y[int(i[0]):int(i[1])], sr=sr, n_mfcc=40))