| At present,the main mode of education in China is exam-oriented education with offline classroom teaching as the core.How to use intelligent way to assist teaching has become the current rese.focus.Classroom interaction data includes a large number of explicit data such as teachers’ courseware,blackboard writing and students’ information,and implicit data such as discourse,gender and emotion generated by teacher-student interaction in the teaching process.These data are the most basic typical characteristics of teaching activities and reflect the essence of teaching to a certain extent.Therefore,multi-dimensional data analysis of classroom interaction is very important.However,the intelligent analysis method adopted by the vast majority of scholars is still to improve S-T analysis method and FIAS method,or to limit the premise of behavior analysis to specific smart classroom environment and online teaching forum,which can not be generally applied to the most important offline classroom teaching environment.At the same time,in the online classroom teaching environment,environmental noise has a wide range of sources,which can not be avoided and has a certain impact on the accuracy of speech recognition,so it is necessary to de noise the classroom voice files.In this paper,the research background of the class analysis is introduced first,and explains the significance of multi-dimensional classroom interaction analysis based on speech recognition,the theory of speech noise reduction,speaker recognition,speech emotion recognition,speech gender recognition used in this paper are also expounded.At the same time,the theories of speech noise reduction,speaker recognition,speech emotion recognition,speech gender recognition,class analysis are also discussed.Secondly,this paper innovatively applies the social network method to the offline classroom,and uses the speaker identity,gender,emotional data obtained above to realize the visualization of offline classroom interaction behavior data through the social network analysis method.The interaction structure(balanced structure,scattered structure,centralized structure)is determined by the interaction density and network diameter.Through the conversion rate of classroom emotions,classroom excitement to determine the level of classroom emotions(high,stable,low),to achieve multi-dimensional qualitative analysis of the classroom.Combined with the above results,through the analysis of individual interaction differences between teachers and students of different genders and identities,classroom interaction structure and classroom mode,teachers can timely grasp the classroom situation and make corresponding adjustments.Finally,a specific case is used to verify the proposed multi-dimensional classroom interaction behavior analysis method based on speech recognition.By comparing individual cases and several cases respectively,the rationality of the multi-dimensional classroom Interactive behavior analysis on the basis of speech recognition is proved,and the research content,innovation and problem points of this paper are summarized. |
