The characteristics of teacher-student verbal interaction behavior in elementary school basic teaching will affect the teaching effect to a certain extent, and this paper uses speech recognition technology to technically empower the analysis of classroom interaction behavior. The ResNet-34 model is used for speech subject recognition, and the loss function AAM-Softmax is introduced to obtain highly distinguishable angular features. With the help of the amplitude spectrum and phase spectrum of the signal as the feature values, the model recognition ability is improved. Taking the elementary school classroom interaction between teachers and students as the research object, based on the iFIAS method for observation, the classroom observation results are analyzed. A teacher’s classroom 2 increased the total number of positive integration zone times by 6 compared to classroom 1, and the number of defective frames decreased to 6. At the same time, the number of steady-state zones (10, 10) reached 105, and the frequency of students’ discussion was effectively increased. In the practical post-test, each class contained 2~3 higher-order thinking skills, and the proportion of students’ practical ability to master the ability to find problems increased by 58%, which shows that the technology-enabled classroom interactive teaching mode can promote the development of students’ higher-order thinking.
