YuanGongND / gopt

Code for the ICASSP 2022 paper "Transformer-Based Multi-Aspect Multi-Granularity Non-native English Speaker Pronunciation Assessment".
BSD 3-Clause "New" or "Revised" License
148 stars 27 forks source link

测试自己的数据时,均值和方差怎么算出来的 #27

Open wangyanquan opened 1 year ago

wangyanquan commented 1 year ago
    # normalize the input to 0 mean and unit std.
    if am=='librispeech':
        dir='seq_data_librispeech'
        norm_mean, norm_std = 3.203, 4.045
    elif am=='paiia':
        dir='seq_data_paiia'
        norm_mean, norm_std = -0.652, 9.737
    elif am=='paiib':
        dir='seq_data_paiib'
        norm_mean, norm_std = -0.516, 9.247
    else:
        raise ValueError('Acoustic Model Unrecognized.')
YuanGongND commented 1 year ago

You should use the mean/std of the acoustic model you use.

librispeech, paiia, paiib are not datasets, but the names of acoustic models.

-Yuan

wangyanquan commented 1 year ago

谢谢大佬

wangyanquan commented 1 year ago

You should use the mean/std of the acoustic model you use.

librispeech, paiia, paiib are not datasets, but the names of acoustic models.

-Yuan

那这个数据在声学模型的哪一个部分可以看到