The information age is an era of innovation. Instructional design and development tools make it easier and faster for teachers to organize teaching content; instructional evaluation tools make it more scientific and objective for teachers to evaluate teaching effects; and student management tools pay attention to individual differences of students and realize diversified evaluation perspectives for students. On the basis of the MIDI dataset, signal separation information filtering and other related operations are used to preprocess ethnic instrumental music signals, followed by the construction of an automatic composition model using deep learning technology. Based on the diversified teaching needs of ethnic instrumental music courses in colleges and universities, an instrumental music teaching system was designed by using the automatic composition model and computer programming language, and an empirical analysis of the informatization teaching of ethnic instrumental music courses in colleges and universities was carried out by using experimental comparative analysis. The results show that the overall level of ethnic music in the experimental group improved significantly P<0.05, indicating that the students in the experimental group made significant progress in ethnic music literacy. In contrast, the students in the control group improved but did not highlight, indicating that the proposed teaching system has a facilitating effect on the enhancement of the teaching performance of ethnic instrumental music courses. This study improves the teaching quality of ethnic instrumental music courses in colleges and universities and is of great significance in enhancing students’ music literacy and skills.