The advancement of digital intelligence technologies presents new opportunities for business English education. Their integration not only enriches course content but also introduces novel teaching methods such as online learning platforms and virtual reality. To further enhance teaching effectiveness, this paper designs a deep learning-based student behavior recognition and teaching effectiveness evaluation mechanism (LBREM). This mechanism first employs deep learning algorithms to analyze student behavior, then evaluates teaching effectiveness based on this analysis. Instructors adjust teaching approaches according to the evaluation outcomes, ultimately improving business English learning outcomes. The object detection model (YOLOv5s-Ghost-D4) and human keypoint detection model (AlphaPose) used in this mechanism demonstrate high recognition performance. When applying the LBREM method for learning effectiveness evaluation, the results show minimal deviation from manual evaluations. Based on the digital-intelligent teaching model and LBREM method, student learning outcomes showed significant improvement: business vocabulary size increased by 155.2%, oral fluency by 26.2%, listening comprehension rate by 29%, writing proficiency by 18.7%, and business adaptability by 20.8%.
