search

yxlu-0102 / MP-SENet

MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectra
MIT License
267 stars 40 forks source link

Result is bad #14

Closed AINoise closed 6 months ago

AINoise commented 6 months ago

when I used " python inference.py --checkpoint_file best_ckpt/g_best " for tested my noisy.wav, The result.wav is bad .The noise is still preserved. Did I make a mistake in my steps？

yxlu-0102 commented 6 months ago

Is your noisy speech data from your own dataset or from the VoiceBank+DEMAND dataset? Does it at a sampling rate of 16 kHz?

AINoise commented 6 months ago

Is your noisy speech data from your own dataset or from the VoiceBank+DEMAND dataset? Does it at a sampling rate of 16 kHz?

It's my dataset and 16kHz. But universality is required. Does it rely strongly on training data?

yxlu-0102 commented 6 months ago

The universality is also related to the dataset size. Our model was trained on the Voicebank+DEMAND dataset, which is not that big. You can fine-tune our model with your data for better performance.