Application of Multisource Data Fusion Analysis in College Vocal Music Teaching

Abstract

This paper constructs a two-level data fusion model with the classroom environment monitoring background of colleges and universities. By judging the validity of the data received by each sensor, the model eliminates the influence of neglected monitoring values on fusion accuracy. It uses the adaptive weighted average method to fuse the data of the same type of sensors in each area and then uses the BP neural network to fuse the heterogeneous sensor data in the area. After each region sends the fusion result to the gateway node, the second-level fusion is performed. In the second-level fusion, the error between the actual output result and the expected output of the BP neural network is calculated, and it is used as the basic probability assignment in DS; then the D-S synthesis rule is used for decision-level fusion so as to realize the integration of the college classroom environment. Aiming at the shortcomings of the D-S evidence theory, we improved the algorithm with respect to the distance of evidence and the conflict factor. Through the exploration of the multisource data fusion analysis method, it is found that it plays an important role in early vocal music teaching. The quality of early vocal music-teaching teachers has a particularly important impact on the music level of college students. Schools with preschool education majors are important bases for teacher training in early vocal music teaching. In order to expand the application scope of the concept of multisource data fusion in the field of vocal music teaching in China, it is necessary to change the traditional vocal music teaching mode of existing college teachers and pay attention to the integration of multisource data fusion analysis methods and the reforms of colleges and universities. In this paper, the effectiveness of the two-level fusion model is verified by using two evaluation indicators: the mean absolute percentage error and the correlation coefficient. Then, comparing the calculation results of the improved algorithm and the classical algorithm in this paper, it is proved that the probability accumulation of this algorithm is more obvious and consistent with the expected results, which shows the optimization effect of the improved method.