This paper explores the optimization methods of higher education management work to develop innovation competence. Focusing on classroom management, a method for assessing learning engagement based on information fusion technology is proposed. The facial images are localized based on Haar classification, and CNN is used to determine the learning emotion. Then, the fatigue level and learning concentration are detected based on eye state recognition and face posture estimation, respectively. The assessment of learning engagement is complete when the multidimensional information of classroom learning is fused. Among them, the better-performing face position decision layer features (F1=76.25%) and the facial expression decision layer features (F1=79.12%) are spliced and fused to increase the classification accuracy. The evaluation of learning engagement aids in improving the quality of classroom management and optimizing educational management of universities.