How to use Gammatone IRM to synthesis the original signal? about speech_feature_extractor HOT 4 CLOSED

是不是应该使用伽马通域的幅值谱*掩蔽值irm呢？

from speech_feature_extractor.

我在Synthesis_speech.py​​中使用函数Synthesis_speech来合成原始信号。
当我得到形状为（64，frame_num）的IRM时，python文件中的代码
masked_abs = np.abs(stft[:win_len//2+1]) * mk[:, i]
使用stft乘以掩码。它与该轴不匹配。因此，python文件仅用于合成由stft提取的那些蒙版，而不用于合成Gammatone过滤器吗？

My problem is the same as yours. How do you solve it?

from speech_feature_extractor.

我也是，怎么弄

from speech_feature_extractor.

Please use the code in speech_utils.py and feature_extractor.py. The rest files are used when I developed this project.
In speech_utils.py, coch_synthesis is used for the maskes defined on the cochleagram (extracted with gammatone filterbanks). While, synthesis_speech is used for the makes defined on the STFT domain.

from speech_feature_extractor.