Design And Realization Of Chinese Reading Grading System Based On Multi-feature Fusion

In recent years,with the improvement of China’s international status,a wave of"Chinese fever" has gradually emerged around the world.On the one hand,the teaching of Chinese as a foreign language is spread through traditional offline Chinese teaching institutions represented by Confucius Institutes.On the other hand,it also needs the support of Chinese learning tools and platforms.However,most of the learning tools and platforms on the market are aimed at Chinese people,and there are few learning platforms that combine Chinese culture with conduct oral training and Chinese follow-up scoring for foreigners.Therefore,in view of the shortcomings of traditional offline Chinese teaching and the lack of oral training for foreign learners,this paper analyzes the limitations of existing voice feature parameter scoring,designs and implements a multi-feature fusion Chinese follow-up scoring system,and obtains the following achievements through the research:First,the overall design scheme of the Chinese follow-up scoring system is explained.Based on the demand analysis of the Chinese follow-up scoring system,the overall structure of the system is determined;the functional modules of the student subsystem and the administrator subsystem are designed;the system database design is completed according to the functional requirements of the system.Second,on the basis of extracting the three traditional acoustic features of pitch trajectory,formant trajectory and Mel frequency cepstral coefficient(MFCC),deep learning features are extracted through wavelet convolution kernel neural network,and a multi-feature fusion is constructed.Chinese follow-up scoring model.Combining the four extracted speech features,the dynamic time warping algorithm is used to perform pattern matching,and the differential evolution algorithm is used to determine the weight ratio of each feature score,and a multi-feature fusion Chinese follow-up scoring algorithm is designed.Then,based on front-end and back-end technologies such as Thymeleaf template engine,SpringBoot framework and MySQL database,a multi-feature fusion Chinese follow-up scoring system is realized.This system is divided into two parts:the student subsystem and the administrator subsystem.Among them,the functions realized by the student subsystem include registration and login module,personal center module,course retrieval and selection module,follow-up learning module and follow-up scoring module,etc.The functions realized by the administrator subsystem include student management module,course management module and Test question management module.Finally,the multi-feature fusion scoring algorithm and the overall system are tested.according to the test results,compared with the traditional convolutional neural network,the wavelet convolutional kernel neural network reduces the network model parameters and accelerates the training speed of the model in the extraction of deep learning features.The correlation of the multi-feature scoring algorithm incorporating wavelet convolution kernel neural network features is 0.0837 higher than that of the single MFCC scoring algorithm,and 0.0345 higher than the three traditional acoustic feature scoring algorithms,which verifies the feasibility and effectiveness of the scoring algorithm.In addition,the system runs stably,and has good compatibility.Each functional module basically achieves the expected design goals.The scoring results are relatively scientific,and it has good application prospects and promotion value.