Research And Implementation Of Speech Evaluation Model Based On Fuzzy Comprehensive Evaluation

With the rapid development of mobile internet,the technology of speech evaluation has become hotspots in the research field of spoken language learning both at home and abroad,it is mainly used in spoken English test and oral English teaching at present.In the process of foreign language learning,Oral learning has always been the weakness of many English learners,so how to make use of speech evaluation technology effectively to solve the problem that the oral English is not up to standard has become a key and difficult point of the current research.This paper based on the analysis of the characteristics of phonological rhythm and the characteristics of English pronunciation focus on the problems that English words evaluation in English learning and puts forward the method of phonological evaluation based on fuzzy comprehensive evaluation.At the same time,this method is transplanted and the speech evaluation based on Android platform is realized,the main contents are as follows:Firstly,the Standard Speech Evaluation Library based on cloud platform is established.The British and American National Corpus is used as the standard corpus.with the methods through the online collection to capture the relevant classification of voice files and saved to the cloud platform.also through the offline collection method and collected 10 people(different gender)read 300 single vowels in 4 cases(normal,rising,falling,wrong read),300 double vowels,600 tone multiple 7 cases(normal,fast reading,slow reading,rise,flat,part fault,fault)a total of 75000 voice file data as the training speech database and classified these English words.The 10 English experts to evaluate according to the pronunciation and build the expert evaluation library.Through the Android platform voice upload module and voice evaluation module to expand the cloud platform voice test library.Secondly,based on the analysis of English words pronunciation characteristics such as the pitch,duration and light stress to extract prosodic features of speech(Mel frequency cepstrum coefficients characteristics,fundamental frequency characteristics and short-time energy characteristics).According to the analytical hierarchy process model,we calculate the weight value of each feature to the final result.The dynamic time warping algorithm are used to compare the standard speech and the test speech,At the same time,a speech evaluation model based on fuzzy comprehensive evaluation is proposed and the speech evaluation score is calculated.Through the Matlab simulation experiments,the performance of the improved speech evaluation algorithm is analyzed and compared.the experiments results show that the improved model has strong efficient and practical.Finally,the improved speech evaluation algorithm is transplanted to the Android platform.Design and implement a voice evaluation system based on cloud platform,through the experimental staff to evaluate the effectiveness of the system test,the test results can basically meet the requirements of voice evaluation.