Quietness has an important influence and regulation on students’ attention, memory, thinking and other cognitive behaviours. Accurate identification of students’ emotions is the basis for establishing harmonious emotional interactions in an intelligent learning environment, and is also an important means of judging learner learning status. This article is based on the theory of deep learning and studies the recognition of emotional states of English learners. Then, training and experiments were conducted on a large-scale learner emotional database built independently. Experiments have proved that this method can quickly and accurately identify students’ emotions. On this basis, the research results of this project can be used in the construction of intelligent learning environments, providing technical support for improving students’ learning mode, promoting students’ emotional interaction, and tapping students’ learning behaviour.