The performance of speaker verification systems is often compromised under real-world environments. For example, variations in handset characteristics could cause severe performance degradation. This paper presents a novel method to overcome this problem by using a non-linear handset mapper. Under this method, a mapper is constructed by training an elliptical basis function network using distorted speech features as inputs and the corresponding clean features as the desired outputs. During feature recuperation, clean features are recovered by feeding the distorted features to the feature mapper. The recovered features are then presented to a speaker model as if they were derived from clean speech. Experimental evaluations based on 258 speakers of the TIMIT and NTIMIT corpuses suggest that the feature mappers improve the verification performance remarkably.

A Optimal M i of Neural NetworkControl Bawd on Fuzzy Logic -yWP Jwsfa%= ~C m D e p m e n t o f~u l l r v e a S l t y ,~~ ~bsbad: The paper presents an integrated design method of neural network control, which is based on the f k z y logic and FCSP(Fuzy Constraint Satisfaction Problem). The whole procedure is discussed and the proposed methodology will be applied to the study of the neurocontrol of a factory boiler model as an illustration.Keywo*

EBF networks are an extension of radial basis function (RBF) networks. Selecting an appropriate number of clusters is a problem in whether RBF or EBF networks. The rival penalized competitive learning (RPCL) algorithm is designed to solve this problem. But its per%oxmance is not satisfactory when the data has overlapped clusters and the input vectors contain dependent components. This paper addresses this problem by in~~rporating full covariance matrices into the original RPCL algorithm. The resulting algorithm, refmed to as the improved RPCL algorithm, progressively eliminates the units whose clusters contain only a small portion of the training data. The improved algorithm is applied to optimize the architecture of elliptical basis !imction networks for process control. The results show that the covariance matrices in the improved RPCL algorithm have a better representation of the clusters.