Objective speech quality is difficult to be measured without the input reference speech.Mapping methods using data mining are investigated and designed to improve the output-based speech quality assessment algorithm.T...Objective speech quality is difficult to be measured without the input reference speech.Mapping methods using data mining are investigated and designed to improve the output-based speech quality assessment algorithm.The degraded speech is firstly separated into three classes(unvoiced,voiced and silence),and then the consistency measurement between the degraded speech signal and the pre-trained reference model for each class is calculated and mapped to an objective speech quality score using data mining.Fuzzy Gaussian mixture model(GMM)is used to generate the artificial reference model trained on perceptual linear predictive(PLP)features.The mean opinion score(MOS)mapping methods including multivariate non-linear regression(MNLR),fuzzy neural network(FNN)and support vector regression(SVR)are designed and compared with the standard ITU-T P.563 method.Experimental results show that the assessment methods with data mining perform better than ITU-T P.563.Moreover,FNN and SVR are more efficient than MNLR,and FNN performs best with 14.50% increase in the correlation coefficient and 32.76% decrease in the root-mean-square MOS error.展开更多
A method utilizing variable depth increments during incremental forming was proposed and then optimized based on numerical simulation and intelligent algorithm.Initially,a finite element method(FEM) model was set up a...A method utilizing variable depth increments during incremental forming was proposed and then optimized based on numerical simulation and intelligent algorithm.Initially,a finite element method(FEM) model was set up and then experimentally verified.And the relation between depth increment and the minimum thickness tmin as well as its location was analyzed through the FEM model.Afterwards,the variation of depth increments was defined.The designed part was divided into three areas according to the main deformation mechanism,with Di(i=1,2) representing the two dividing locations.And three different values of depth increment,Δzi(i=1,2,3) were utilized for the three areas,respectively.Additionally,an orthogonal test was established to research the relation between the five process parameters(D and Δz) and tmin as well as its location.The result shows that Δz2 has the most significant influence on the thickness distribution for the corresponding area is the largest one.Finally,a single evaluating indicator,taking into account of both tmin and its location,was formatted with a linear weighted model.And the process parameters were optimized through a genetic algorithm integrated with an artificial neural network based on the evaluating index.The result shows that the proposed algorithm is satisfactory for the optimization of variable depth increment.展开更多
基金Projects(61001188,1161140319)supported by the National Natural Science Foundation of ChinaProject(2012ZX03001034)supported by the National Science and Technology Major ProjectProject(YETP1202)supported by Beijing Higher Education Young Elite Teacher Project,China
文摘Objective speech quality is difficult to be measured without the input reference speech.Mapping methods using data mining are investigated and designed to improve the output-based speech quality assessment algorithm.The degraded speech is firstly separated into three classes(unvoiced,voiced and silence),and then the consistency measurement between the degraded speech signal and the pre-trained reference model for each class is calculated and mapped to an objective speech quality score using data mining.Fuzzy Gaussian mixture model(GMM)is used to generate the artificial reference model trained on perceptual linear predictive(PLP)features.The mean opinion score(MOS)mapping methods including multivariate non-linear regression(MNLR),fuzzy neural network(FNN)and support vector regression(SVR)are designed and compared with the standard ITU-T P.563 method.Experimental results show that the assessment methods with data mining perform better than ITU-T P.563.Moreover,FNN and SVR are more efficient than MNLR,and FNN performs best with 14.50% increase in the correlation coefficient and 32.76% decrease in the root-mean-square MOS error.
文摘A method utilizing variable depth increments during incremental forming was proposed and then optimized based on numerical simulation and intelligent algorithm.Initially,a finite element method(FEM) model was set up and then experimentally verified.And the relation between depth increment and the minimum thickness tmin as well as its location was analyzed through the FEM model.Afterwards,the variation of depth increments was defined.The designed part was divided into three areas according to the main deformation mechanism,with Di(i=1,2) representing the two dividing locations.And three different values of depth increment,Δzi(i=1,2,3) were utilized for the three areas,respectively.Additionally,an orthogonal test was established to research the relation between the five process parameters(D and Δz) and tmin as well as its location.The result shows that Δz2 has the most significant influence on the thickness distribution for the corresponding area is the largest one.Finally,a single evaluating indicator,taking into account of both tmin and its location,was formatted with a linear weighted model.And the process parameters were optimized through a genetic algorithm integrated with an artificial neural network based on the evaluating index.The result shows that the proposed algorithm is satisfactory for the optimization of variable depth increment.
