Intelligent Teaching Behavior Analysis System Based On Speaker Classification Algorithm

In the era of Educational Informatization 2.0,digital technology is increasingly permeating teaching activities,but there is still a lack of intelligent methods in teaching evaluation.Traditional evaluation methods make teaching evaluation activities lack timeliness,convenience and objectivity.And daily,continuous teaching analysis and evaluation will help teachers to carry out in-depth teaching reflection,which is conducive to the development of teachers' professional quality,and then improve the quality of teaching.Under the guidance of teaching theory,the combination of cutting-edge artificial intelligence technology and traditional evaluation methods is not only in line with the goals of education information technology to improve the professional ability of teachers and promote teaching quality,but also conducive to the large sample teaching analysis in teaching research activities,Can promote the development of teaching theory.Discourse is the main carrier of knowledge transmission in classroom teaching activities,and the turn-turning between teachers and students who are the main speakers in the classroom constitutes the process of imparting knowledge.Based on the characteristics of classroom discourse,this paper proposes an unsupervised speaker identity classification algorithm based on pre-clustering.Based on the S-T analysis method in classroom teaching analysis,taking classroom recording as input,this algorithm is used to derive the ST curve of classroom teaching.And classroom discourse time-sharing map,design and implement an intelligent and visual classroom teaching evaluation system.The main work and innovation of this article are as follows:First,comparatively analyze the three commonly used teaching behavior analysis methods,combined with the technical research foundation of this study-speaker classification algorithm,and determine the theoretical basis of the S-T analysis method as an intelligent teaching behavior analysis system.Secondly,the database of speaker classification algorithm research lacks classroom teaching scenes.Based on real classroom video data,this paper organizes and builds a classroom teaching conversation speech database,and analyzes the characteristics of speech signals in classroom teaching scenes from the perspective of acoustic characteristics.Thirdly,according to the characteristics of unbalanced distribution of speaker utterances in classroom utterances and teachers as the single main speaker,a dual clustering speaker identity classification algorithm is proposed.The algorithm firstly determines part of the main speaker's speech segments through pre-hierarchical clustering,then uses the clustered speech to build a teacher utterance model,divides the teacher utterance part by one-by-one recognition,and finally uses hierarchical clustering to speak the remaining speech People classification.And through experiments verified that the classification algorithm proposed in this study has a higher recognition effect.Finally,the algorithm proposed in this paper and the ST analysis method are combined to design and implement an intelligent teaching behavior analysis system.Classroom video or audio recording is used as input to determine the teaching mode.3.The teaching model structure distribution map is presented in three ways,and the effectiveness of the system is verified through the comparison experiment with the traditional ST analysis method.