Research On Speech Performance Evaluation Method Of Hearing Aid Based On MARS

With the society’s aging progress,the amount of hearing patients grows greatly.As the main hearing improving device,its sound quality is highly related to the health of hearing patients.So it is important to objectively evaluate the sound quality of hearing aids accurately.Taking advantage of the large data produced in medical process is challenging and promising.This thesis research on an objective evaluation method of speech quality based on Multivariate adaptive regression spline(MARS),and its improvement;AP cluster algorithm and its improvement.And finally proposed a more effective algorithm based on the above two methods.Simulation experiments are done for verification.The main research work and innovations of the thesis are as follows:(1)A test platform for the objective assessment system of hearing aids was built,based on which the hearing aid voice database was built.The features of hearing aid’s output speech signals are extracted and implemented using matlab.(2)An algorithm based on MARS for objective assessment of hearing aids is proposed.In the algorithm,features are optimized firstly,then the objective evaluation model of the hearing aid is established,which gives out the objective score.An evaluation index r about accuracy considering fault-tolerant error is proposed.With other three traditional indicators,the performance of algorithm is verified,and the outcome shows that the algorithm is effective.(3)A new clustering algorithm-semi-supervised AP based on density path(SMAP)is proposed.The algorithm firstly semi-supervised the unknown sample data,and then based on the density path,the traditional is improved.AP similarity"Euclidean distance" and obtained the clustering model.By analyzing the experimental simulation results,the algorithm performance evaluation indicators are summarized,the improved algorithm is valid and feasible,and the accuracy is high.(4)An improved objective evaluation method for speech quality of hearing aids-SPMARS algorithm based on the clustering analysis is proposed.In the algorithm,the speech data are divided into different clusters,and then MARS objective evaluation on different data clusters are done respectively.Through the simulation experiment and using evaluation index,the improved algorithm is verified,which proves that the algorithm can accurately evaluate the speech quality of hearing aids.