A design method for noise canceller using parallel recurrent neural filters <Abstract> This paper presents a spectral subtraction using the classifications between the speech dominant and the noise one. In our system, a new classification scheme between the speech dominant and the noise one is proposed. The proposed classifications use the standard deviation of the spectrum of observation signal in each critical band. We have introduced two This paper presents a design method for noise canceller using parallel recurrent neural filters (PRNFs). PRNFs consist of several small-scale recurrent neural filters (RNFs) connected in parallel to reduce calculation for high-speed processing. RNF is a neural filter using the recurrent neural network that has self-feedback at hidden neurons. In addition, the value of learning coefficient is varied dynamically to accelerate PRNFs. Finally, we show the results of the numerical simulation and illustrate the effectiveness of the proposed method for nonlinear path. |