53 lines
2.3 KiB
Python
53 lines
2.3 KiB
Python
|
import argparse
|
||
|
|
import functools
|
||
|
|
import itertools
|
||
|
|
|
||
|
|
from mvector.predict import MVectorPredictor
|
||
|
|
from mvector.utils.utils import add_arguments, print_arguments
|
||
|
|
from mvector.data_utils.audio import AudioSegment
|
||
|
|
|
||
|
|
parser = argparse.ArgumentParser(description=__doc__)
|
||
|
|
add_arg = functools.partial(add_arguments, argparser=parser)
|
||
|
|
add_arg('configs', str, 'configs/ecapa_tdnn.yml', '配置文件')
|
||
|
|
add_arg('use_gpu', bool, True, '是否使用GPU预测')
|
||
|
|
add_arg('audio_path1', str, 'dataset/source/王翔/wx1_5.wav', '预测第一个音频')
|
||
|
|
add_arg('audio_path2', str, 'dataset/source/刘云杰/lyj_no_5.wav', '预测第二个音频')
|
||
|
|
add_arg('threshold', float, 0.7, '判断是否为同一个人的阈值')
|
||
|
|
add_arg('model_path', str, 'models/test_model', '导出的预测模型文件路径')
|
||
|
|
args = parser.parse_args()
|
||
|
|
# print_arguments(args=args)
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
# 获取识别器
|
||
|
|
predictor = MVectorPredictor(configs=args.configs,
|
||
|
|
model_path=args.model_path,
|
||
|
|
use_gpu=args.use_gpu)
|
||
|
|
|
||
|
|
def load_audio_paths(file_path):
|
||
|
|
with open(file_path, 'r',encoding='utf-8') as file:
|
||
|
|
return [line.strip() for line in file if line.strip()]
|
||
|
|
|
||
|
|
|
||
|
|
def compare_audio_files(audio_paths, threshold):
|
||
|
|
# itertools.combinations 生成所有可能的两两组合，无重复
|
||
|
|
for audio1, audio2 in itertools.combinations(audio_paths, 2):
|
||
|
|
dist = predictor.contrast(audio1, audio2)
|
||
|
|
if dist > threshold:
|
||
|
|
print(f"{audio1} 和 {audio2} 为同一个人，相似度为：{dist}")
|
||
|
|
# else:
|
||
|
|
# print(f"{audio1} 和 {audio2} 不是同一个人，相似度为：{dist}")
|
||
|
|
|
||
|
|
file_path = 'dataset/ces.txt' # 假设音频路径存储在此文件中
|
||
|
|
|
||
|
|
# 执行比对
|
||
|
|
audio_paths = load_audio_paths(file_path)
|
||
|
|
compare_audio_files(audio_paths, args.threshold)
|
||
|
|
# # AudioSegment.silent_semoval(args.audio_path1, args.audio_path1)
|
||
|
|
# # AudioSegment.silent_semoval(args.audio_path2, args.audio_path2)
|
||
|
|
# dist = predictor.contrast(args.audio_path1, args.audio_path2)
|
||
|
|
# if dist > args.threshold:
|
||
|
|
# print(f"{args.audio_path1} 和 {args.audio_path2} 为同一个人，相似度为：{dist}")
|
||
|
|
# else:
|
||
|
|
# print(f"{args.audio_path1} 和 {args.audio_path2} 不是同一个人，相似度为：{dist}")
|