In the era of smart education, online courses as the avant-garde force in the educational field are leading the way in innovating teaching methods. Although online learning platforms provide students with convenient channels for learning, issues such as course quality, personalized service, and learning motivation still exist. This study, based on China University MOOC, proposes a personalized online course recommendation method based on emotion recognition, aimed at deeply understanding students’ emotional states to enhance the accuracy and personalization of course recommendations. Initially, this paper collected a dataset of course user comments from China University MOOC and built an emotional dictionary in the education domain to analyze users’ emotional states. Combining emotion analysis, user characteristics, and course features, the SAFM and SDFM models were proposed, incorporating a negative sampling method to generate personalized course recommendations. The experiments prove that this method effectively enhances students’ learning motivation and participation, offering new insights for the development of online education platforms.