From: A recursive expectation-maximization algorithm for speaker tracking and separation
Algorithm | |||||||||
---|---|---|---|---|---|---|---|---|---|
Proposed | Baseline | ||||||||
SNR | SIR(Input) | \(\textrm {SIR}_{1}(\hat {\mathbf {D}})\) | \(\textrm {SIR}_{1}(\hat {\mathbf {G}})\) | \(\textrm {SIR}_{2}(\hat {\mathbf {D}})\) | \(\textrm {SIR}_{2}(\hat {\mathbf {G}})\) | \(\textrm {SIR}_{1}(\hat {\mathbf {D}})\) | \(\textrm {SIR}_{1}(\hat {\mathbf {G}})\) | \(\textrm {SIR}_{2}(\hat {\mathbf {D}})\) | \(\textrm {SIR}_{2}(\hat {\mathbf {G}})\) |
40 | 2.96 | 9.53 | 10.79 | 4.72 | 9.57 | 3.39 | 5.24 | 6.25 | 6.67 |
30 | 2.96 | 9.55 | 10.84 | 4.72 | 9.52 | 3.33 | 5.12 | 6.14 | 6.27 |
20 | 2.96 | 9.46 | 10.40 | 4.57 | 9.45 | 3.37 | 4.79 | 5.72 | 5.92 |
10 | 2.96 | 9.55 | 11.04 | 4.70 | 9.01 | 3.42 | 4.17 | 5.47 | 5.08 |
0 | 2.96 | 9.17 | 10.19 | 4.57 | 7.87 | 3.52 | 2.19 | 5.45 | 3.19 |