Kento Ohtani, "Sound Source Enhancement of Monaural Music with Classfication Based SNR Estimation"

2016/09/13 22:38 に 後藤太一 が投稿

In this report, he propose a monaural source separation method that estimates SNRs of source signals using RNN.
Experiment using 100 music used in SiSEC 2015 shows that  proposed method shows highest SDRs.
But Separation quality is not enough.
So as a future works, there are Quantization method of SNR of source signals, Post-processing and RNN structure.