complex-beta-nmf-master
所属分类:自然语言处理
开发工具:matlab
文件大小:16KB
下载次数:2
上传日期:2019-12-27 10:24:18
上 传 者:
呼噜噜啦
说明: 复杂的非负矩阵分解代码,实现更好的效果来分解
(Complex non negative matrix decomposition code for better results)
文件列表:
LICENSE (1816, 2018-04-27)
audio_files (0, 2018-04-27)
audio_files\.dummy.txt (40, 2018-04-27)
dataset (0, 2018-04-27)
dataset\.dummy.txt (40, 2018-04-27)
functions (0, 2018-04-27)
functions\GetSDR.m (1775, 2018-04-27)
functions\NMF.m (2188, 2018-04-27)
functions\STFT.m (1033, 2018-04-27)
functions\cnmf_wisa.m (2381, 2018-04-27)
functions\complex_ISNMF_unconstrained.m (3847, 2018-04-27)
functions\complex_betaNMF.m (3101, 2018-04-27)
functions\get_data_DSD_cbetanmf.m (1371, 2018-04-27)
functions\iSTFT.m (1501, 2018-04-27)
functions\wiener_nmf.m (397, 2018-04-27)
metrics (0, 2018-04-27)
metrics\.dummy.txt (40, 2018-04-27)
scripts (0, 2018-04-27)
scripts\compute_score_cbetanmf.m (1543, 2018-04-27)
scripts\learning_beta_cbnmf.m (2823, 2018-04-27)
scripts\separation_cbetanmf.m (2407, 2018-04-27)
scripts\set_settings_cbetanmf.m (458, 2018-04-27)
# Complex NMF with the beta-divergence GitHub repository
Code for complex NMF with the beta-divergence
Here, you will find the code related to complex NMF with the beta-divergence. It is applied to the task of harmonic/percussive source separation.
If you use any of the things existing in this repository, please cite the [corresponding paper](https://hal.archives-ouvertes.fr/hal-017796***).
### Dataset set-up
To reproduce the experiments conducted in our paper, you will need to download the [Dexmixing Secret Database (DSD100)](http://www.sisec17.audiolabs-erlangen.de) and to place its content in the `dataset`.
If you use this dataset, you will end up with the proper directory structure and file names, as used in the `functions/get_data_DSD_cbetanmf` function.
If you want to use a different dataset, then you have two options:
- either you format your file names and directory structure to match the one from DSD100;
- or you modify the file reading function `get_data_DSD_cbetanmf` to suit your needs.
### Experiments
The two main scripts to reproduce the experiments are in the `scripts` folder:
`learning_beta_cbnmf` will perform a grid search for beta and plot the separation quality (SDR, SIR and SAR) as a function of this parameter. This apply for Complex NMF and NMF.
`separation_cbetanmf` will perform the benchmark conducted in the paper, that is, comparing NMF and complex NMF with Euclidean, IS and beta divergences. The audio files will be record in the `audio_files` folder and the script `comopute_score_cbetanmf` computes the SDR, SIR and SAR for this benchmark.
近期下载者:
相关文件:
收藏者: