Abstract: The “finger movement and dictation” safety confirmation can effectively reduce the risk of operational errors in the production process of coal mines. In order to solve the problem of lack of effective supervision in the implementation of traditional methods,a set of “finger movement and dictation” safety confirmation system in coal mine is designed based on deep learning technology. To judge whether the standardized operation is achieved by testing the cooperation of body and language,and to supervise,standardize and implement the implementation of the “finger movement and dictation” safety confirmation method at the technical level. Firstly, Whisper speech model is used to identify “ dictation”, and OpenPose is used to estimate human posture. Then,in order to improve the detection accuracy,an enhanced Transformer is proposed to improve the YOLOv8 model. The target detection of “finger movement” is carried out. Finally,the designed “ finger movement and dictation” confirmation system is applied to two typical scenarios, namely checking whether personal protective equipment is complete and worn properly,and checking the key instrument equipment and operation conditions at the workplace. The system is tested on datasets constructed in the two scenarios,and the detection accuracy is 93. 86% and 86. 24% respectively,which proves the effectiveness of the system.