The Gaussian mixture model (GMM) method is popular and efficient for voice conversion (VC), but it is often subject to overfitting. In this paper, the principal component regression (PCR) method is adopted for the spectral mapping between source speech and target speech, and the numbers of principal components are adjusted properly to prevent the overfitting. Then, in order to better model the nonlinear relationships between the source speech and target speech, the kernel principal component regression (KPCR) method is also proposed. Moreover, a KPCR combined with GMM method is further proposed to improve the accuracy of conversion. In addition, the discontinuity and oversmoothing problems of the traditional GMM method are also addressed. On the one hand, in order to solve the discontinuity problem, the adaptive median filter is adopted to smooth the posterior probabilities. On the other hand, the two mixture components with higher posterior probabilities for each frame are chosen for VC to reduce the oversmoothing problem. Finally, the objective and subjective experiments are carried out, and the results demonstrate that the proposed approach shows greatly better performance than the GMM method. In the objective tests, the proposed method shows lower cepstral distances and higher identification rates than the GMM method. While in the subjective tests, the proposed method obtains higher scores of preference and perceptual quality.
The aim of the study was to evaluate the possibility of applying different methods of data mining to model the inflow of sewage into the municipal sewage treatment plant. Prediction models were elaborated using methods of support vector machines (SVM), random forests (RF), k-nearest neighbour (k-NN) and of Kernel regression (K). Data consisted of the time series of daily rainfalls, water level measurements in the clarified sewage recipient and the wastewater inflow into the Rzeszow city plant. Results indicate that the best models with one input delayed by 1 day were obtained using the k-NN method while the worst with the K method. For the models with two input variables and one explanatory one the smallest errors were obtained if model inputs were sewage inflow and rainfall data delayed by 1 day and the best fit is provided using RF method while the worst with the K method. In the case of models with three inputs and two explanatory variables, the best results were reported for the SVM and the worst for the K method. In the most of the modelling runs the smallest prediction errors are obtained using the SVM method and the biggest ones with the K method. In the case of the simplest model with one input delayed by 1 day the best results are provided using k-NN method and by the models with two inputs in two modelling runs the RF method appeared as the best.