Performance Measurement in Blind Source Separation
Emmanuel Vincent, Rémi Gribonval and Cédric Févotte
About this page
This page provides sound files used in Section IV of the article "Performance measurement in Blind Audio Source Separation" published in IEEE Transactions on Speech and Audio Processing.
Abstract: In this article, we discuss the evaluation of Blind Audio Source Separation (BASS) algorithms. Depending on the exact application, different distortions can be allowed between an estimated source and the wanted true source. We consider four different sets of such distortions, from time-invariant gains to time-varying filterings. In each case we decompose the estimated source into a true source part plus error terms corresponding to interferences, additive noise and algorithmic artifacts. Then we derive a global performance measure using an energy ratio, plus a separate performance measure for each error term. These measures are computed and discussed on the results of several BASS problems with various difficulty levels.
This page contains three parts, corresponding to the three examples of Sections IVA, IVB and IVC and to the Tables II, III and IV of the article. For each example, the estimated source s1_est is decomposed as a sum of three terms: s_dist, e_interf and e_artif, that we provide as WAV files for listening. We also provide the WAV file of e_total=e_interf+e_artif. The performance criteria are copied from the article. The results for the other estimated sources are not presented for the sake of brevity.
Instantaneous 2x2 mixture
Mixture: x
Sources: s1 s2
Algorithm | Estimated source | Allowed distortion | True source part | Total error | SDR | Interferences | SIR | Artifacts | SAR |
---|---|---|---|---|---|---|---|---|---|
JADE | s1_est | TI Gain | s_dist | e_total | 26 dB | e_interf | 26 dB | e_artif | 75 dB |
TFBSS | s1_est | TI Gain | s_dist | e_total | 53 dB | e_interf | 53 dB | e_artif | 71 dB |
Convolutive 2x2 mixture
Mixture: x
Sources: s1 s2
Algorithm | Estimated source | Allowed distortion | True source part | Total error | SDR | Interferences | SIR | Artifacts | SAR |
---|---|---|---|---|---|---|---|---|---|
FICA | s1_est | TI Gain | s_dist | e_total | -11 dB | e_interf | 7 dB | e_artif | -10 dB |
TI Filt | s_dist | e_total | 6 dB | e_interf | 6 dB | e_artif | 20 dB | ||
TV Filt | s_dist | e_total | 3 dB | e_interf | 8 dB | e_artif | 5 dB | ||
OFICA | s1_est | TI Gain | s_dist | e_total | -11 dB | e_interf | 12 dB | e_artif | -10 dB |
TI Filt | s_dist | e_total | 11 dB | e_interf | 12 dB | e_artif | 21 dB | ||
TV Filt | s_dist | e_total | 5 dB | e_interf | 14 dB | e_artif | 6 dB |
Instantaneous 2x3 mixture
Mixture: x
Sources: s1 s2 s3
Algorithm | Estimated source | Allowed distortion | True source part | Total error | SDR | Interferences | SIR | Artifacts | SAR |
---|---|---|---|---|---|---|---|---|---|
STFTC | s1_est | TI Gain | s_dist | e_total | 2 dB | e_interf | 15 dB | e_artif | 3 dB |
TI Filt | s_dist | e_total | 5 dB | e_interf | 14 dB | e_artif | 6 dB | ||
TV Filt | s_dist | e_total | 6 dB | e_interf | 11 dB | e_artif | 8 dB | ||
MPC | s1_est | TI Gain | s_dist | e_total | 4 dB | e_interf | 19 dB | e_artif | 4 dB |
TV Filt | s_dist | e_total | 5 dB | e_interf | 13 dB | e_artif | 5 dB | ||
TV Filt | s_dist | e_total | 6 dB | e_interf | 14 dB | e_artif | 7 dB |