def vad():
logger.info(vad_sh)
os.system(vad_sh)
if not os.path.exists(segment):
logger.error("vad failed！")
os._exit(0)
else:
logger.info("vad succsee！")
def main(wavspath, outputpath, value, tocut):
result_error = {}
for root, dirs, files in os.walk(wavspath):
for file in files:
if not file.endswith('wav'):
continue
package_name = file.rstrip('.wav')
logger.info(f'处理{package_name}中.')
wav_path = os.path.join(root, package_name + '.wav')
txt_path = os.path.join(root, package_name + '.txt')
wavmaps = read_txttime(txt_path)
wavslist = sorttime(wavmaps, txt_path)
tail_time = read_tail_time(wavslist)
wave_time = read_wavtime(wav_path)
time_ = tail_time - wave_time
diff_time = round(time_, 3)
logger.info(f'时间点差为{diff_time}.')
if not judgement_time(time_, threshold=value):
result_error[package_name] = {'tail_time': tail_time, 'wave_time': wave_time}
continue
if tocut:
outputpath = mkdir(os.path.join(outputpath, package_name))
pool = Pool(processes=8)
logger.info(f'切割 {file} 中...')
for item in wavslist:
wavname = item.get('wavname')
outputwav = os.path.join(outputpath, wavname)
start = item.get('start')
end = item.get('end')
pool.apply_async(wavcut, (wav_path, outputwav, start, end))
pool.close()
pool.join()
logger.info('检测完成！')
if result_error:
logger.error('%s 未通过检测，需手动调整切割 [%s]' % (list(result_error.keys()), result_error))
else:
logger.info('全部音频通过检测！')
def run_decode():
# 数据vad
if not os.path.exists(segment):
vad()
else:
logger.info("sgements file has exist, vad pass!")
# 数据准备
cmd = f"cp -r {segment_dir} data/"
logger.info(cmd)
os.system(cmd)
# 解码操作
logger.info(decode_sh)
os.system(decode_sh)
if os.path.exists(decode_file):
logger.info("####------解码完成!--------####")
else:
logger.error("####------解码失败!--------####")
result_map, whsum = CreatPTFile(project_txts_path, dic_map, classify).run()
save_to_file(save_upfile, result_map)
logger.info("Success!!! 共【%s】包【%s】条音频" %
(len(list(result_map.keys())), whsum))
if uposs:
updirpath, amount = uploadpre(project_root, project_wavs_path,
save_upname)
logger.info("音频准备完成")
if whsum != amount:
logger.error(f"音频数【{amount}】与文本数【{whsum}】不对应，请检查！")
else:
cmd = f"ossutil64 cp -r -u {updirpath} oss://cn-wav-crowd/{save_upname}"
logger.info(cmd)
os.system(cmd)
dirname = f"updir_{save_upname}"
os.chdir(project_root)
cmd = f'ossutil64 ls -s oss://cn-wav-crowd/{save_upname} | grep ".wav" | wc -l'
out, err = exe_command(cmd)
try:
upcount = int(out)
parser.add_argument('-cutwav', type=bool, default=True, help='是否切割音频')
parser.add_argument('-outpath', default=None, help='包含音频文件与对应文本文件的文件夹路径')
parser.add_argument('-sleep_time', type=int, default=2, help='(选填)设置停顿时间')
args = parser.parse_args()
CUTWAV = args.cutwav
WAVSPATH = args.path
OUTPUTPATH = args.outpath
VALUE = args.sleep_time # 录音停顿时间
if OUTPUTPATH == None:
OUTPUTPATH = WAVSPATH
if not os.path.isdir(WAVSPATH):
logger.error('路径不正确，检查后请重试')
sys.exit()
def mkdir(path):
if not os.path.exists(path):
os.makedirs(path)
return path
def sorttime(maps, txt_path):
wavinfo = list()
with open(txt_path, 'w', encoding='utf-8') as fw:
fw.write(maps.pop('header'))
for wavname, values in sorted(maps.items(), key=lambda x: float(x[1].get('start'))):
start = values.get('start')