Table 5 Performance degradation of proposed method for suboptimal parameter settings. We chose λT=0.5=const. and λE=8.0=const. which are optimal for SNR =1 dB and evaluated SNRs − 1 dB, \(\dots \), 3 dB. For each performance value of the proposed method, the degradation relative to optimum parameter settings for λE and λT is given in parentheses. For this experiment, ego-noise by movements of Sequence III is used
From: Motor data-regularized nonnegative matrix factorization for ego-noise suppression
SNR= − 1 dB | SNR=0 dB | SNR= + 1 dB | SNR= + 2 dB | SNR= + 3 dB | ||
|---|---|---|---|---|---|---|
PESQ=1.06 | PESQ=1.07 | PESQ=1.08 | PESQ=1.09 | PESQ=1.1 | ||
NMF | SDR [dB] | 5.88 | 6.08 | 6.34 | 6.55 | 6.72 |
SAR [dB] | 7.88 | 8.14 | 8.48 | 8.76 | 9.00 | |
PESQ | 1.25 | 1.26 | 1.29 | 1.31 | 1.33 | |
Proposed | SDR[dB] | 6.25 (− 0.15) | 6.87 (− 0.1) | 7.46 (± 0.0) | 7.87 (− 0.11) | 8.19 (− 0.22) |
SAR [dB] | 8.41 (− 0.24) | 9.36 (− 0.1) | 9.78 (± 0.0) | 10.19 (− 0.19) | 10.51 (− 0.38) | |
PESQ | 1.26 (− 0.04) | 1.30 (− 0.02) | 1.38 (± 0.0) | 1.37 (− 0.02) | 1.39 (− 0.02) |