Project for the course DT2119 Speech and Speaker Recognition offered by KTH in 2018.
The project in inspired by [1].
Execution instructions:
- Download MIR-1K dataset from https://sites.google.com/site/unvoicedsoundseparation/mir-1k
- Execute create_dataset.py to split and preprocess data
- Execute train.py to train the network
- Execute eval.py to evaluate the network
[1] P.-S. Huang, M. Kim, M. Hasegawa-Johnson, and P. Smaragdis, “Singing-voice separation from monaural recordings using deep recurrent neural networks.,” in ISMIR, 2014, pp. 477–482.