The development of deep learning technology provides new ideas for the assessment of teaching quality in the field of education. In this study, a deep learning-based face recognition emotion detection system is designed and applied to vocational education teaching scenarios. The system analyses the emotional state of students by capturing their facial expressions in real time, providing objective feedback data for teachers. The study uses controlled experimental methods to confirm that the system can significantly improve classroom engagement and learning effects. Teachers adjust their teaching strategies based on the emotional feedback data and accurately grasp the pace of teaching, which effectively solves the problems of strong subjectivity and poor real-time performance of traditional teaching assessment. The study provides an innovative technological path for the improvement of teaching quality in vocational education, and is of great significance in promoting teaching precision and personalisation.