speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

Update test_sv_inference_pipeline.py (#635)

Browse Source 
add unit tests for sv model
This commit is contained in:
Zhihao Du 2023-06-15 22:11:06 +08:00 committed by GitHub
parent 6e9d84e4aa
commit d74fdb3a74
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 13 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
tests/test_sv_inference_pipeline.py
View File

@ -19,30 +19,20 @@ class TestXVectorInferencePipelines(unittest.TestCase):
task=Tasks.speaker_verification, task=Tasks.speaker_verification,
model='damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch' model='damo/speech_xvector_sv-zh-cn-cnceleb-16k-spk3465-pytorch'
) )
# 提取不同句子的说话人嵌入码
rec_result = inference_sv_pipline(
audio_in='https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_enroll.wav')
enroll = rec_result["spk_embedding"]
rec_result = inference_sv_pipline( # the same speaker
audio_in='https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_same.wav') rec_result = inference_sv_pipline(audio_in=(
same = rec_result["spk_embedding"] 'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_enroll.wav',
'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_same.wav'))
rec_result = inference_sv_pipline( assert abs(rec_result["scores"][0]-0.85) < 0.1 and abs(rec_result["scores"][1]-0.14) < 0.1
audio_in='https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_different.wav') logger.info(f"Similarity {rec_result['scores']}")
different = rec_result["spk_embedding"]
# 对相同的说话人计算余弦相似度
sv_threshold = 0.9465
same_cos = np.sum(enroll * same) / (np.linalg.norm(enroll) * np.linalg.norm(same))
same_cos = max(same_cos - sv_threshold, 0.0) / (1.0 - sv_threshold) * 100.0
logger.info("Similarity: {}".format(same_cos))
# 对不同的说话人计算余弦相似度
diff_cos = np.sum(enroll * different) / (np.linalg.norm(enroll) * np.linalg.norm(different))
diff_cos = max(diff_cos - sv_threshold, 0.0) / (1.0 - sv_threshold) * 100.0
logger.info("Similarity: {}".format(diff_cos))
# different speaker
rec_result = inference_sv_pipline(audio_in=(
'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_enroll.wav',
'https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/sv_example_different.wav'))
assert abs(rec_result["scores"][0]-0.0) < 0.1 and abs(rec_result["scores"][1]-1.0) < 0.1
logger.info(f"Similarity {rec_result['scores']}")
if __name__ == '__main__': if __name__ == '__main__':
unittest.main() unittest.main()