Automatic Evaluation System of Spoken English for Multi Person Dialogue in English Teaching based on Multi Feature Fusion

Abstract

The problem of imperfect detection function of the existing automatic evaluation system leads to the system occupying too much memory. This paper designs a multi feature fusion based automatic evaluation system for oral English teaching. In the hardware part, the sample is loaded into the input pin of the module to be tested, and the sensor port and peripheral configuration are connected. In the software part, the frequency domain features of English teaching are extracted to obtain the voice feature parameters of time transformation. The multi feature fusion technology is used to construct the adaptive acoustic model of multi person conversation, calculate the acoustic observation value, set the detection threshold by using the double threshold comparison method, and set the detection function of the automatic evaluation module of the software for deep learning. Experimental results: compared with the two existing automatic evaluation systems, the average memory occupied by the designed automatic evaluation system is 34.868%, 49.057% and 48.752% respectively, which proves that the automatic evaluation system combined with multi feature fusion technology has higher practical application valuet.