Supervised deep learning based speech enhancement

Code of paper Li, H., Xu, Y., Ke, D., & Su, K. (2020). Improving speech enhancement by focusing on smaller values using relative loss. IET Signal Processing, 14(6), 374-384.

all hyperparameter are set in "./nn_se/FLAGS.py", each experiment corresponds to a class in "FLAGS.py".

run OMP_NUM_THREADS=1 python -m xx._1_preprocess to prepare data.

run CUDA_VISIBLE_DEVICES=2 OMP_NUM_THREADS=1 python -m xx._2_train to start training.

run OMP_NUM_THREADS=1 python -m xx._3_eval_metrics 2 to get test results, 2 is the number of processor used for testing, OMP_NUM_THREADS=1 is needed for fast running.

xx is class name (also experiment name, code folder name) setting in FLAGS.py (last line PARAM = xx).

inference interface are gaven in "./nn_se/inference.py".

Please read the code for other content.