The training target is obtained by preserving the first 100 ms reflections

Xiaobin-Rong / gtcrn

The official implementation of GTCRN, an ultra-lite speech enhancement model.

MIT License

217 stars 37 forks source link

Closed zdongchen closed 1 month ago

zdongchen commented 1 month ago

请问论文3.1 Datasets 在数据集制作时 "训练目标是通过保留前 100 毫秒反射来实现的" 这句话是什么意思，不太能理解，望解答，感谢

Xiaobin-Rong commented 1 month ago

通过保留早期混响，让clean可以和含混响的noisy的在时域上对齐。这也是大部分SE文章的做法