Simulated annealing based parameter optimization of Time-Frequency ε-filter utilizing correlation coefficient

Time-Frequency ε-filter (TF ε-filter) can reduce different types of noise from a single-channel noisy signal while preserving the signal that varies drastically such as a speech signal. It can reduce not only small stationary noise but also large nonstationary noise. However, it has some parameters whose values are set empirically. So far, there are few studies to optimize the parameter of TF ε-filter automatically. In this paper, we employ the correlation coefficient of the filter output and the difference between the filter input and output as the evaluation function of the parameter optimization. We also propose an algorithm to set the optimal parameter of TF ε-filter automatically. The experimental results show that we can obtain the adequate parameter in TF ε-filter automatically by using the proposed method.