Detect the electromyographic signal of the typical movement of traditi onal Taijiquan long-term practitioners and short-term practitioners of "knee- kne e and stance",and analyze the changes of the electromyographic signals of the m ain muscles of the upper limbs,lower limbs and trunk during the punching proces s.According to the rules and main points of traditional Taijiquan,the detected EMG signals are denoised,feature values are extracted,and BP neural network m odels are established.The movement patterns of long-term practitioners and sho r t-term practitioners are classified and analyzed.The experimental results show that in the process of long-term practitioners,the left and right sides of the body show the states of "left solid,right virtual" and "right solid,left virtu al",and the limbs show the characteristics of "sections through";Compared with the action recognition rate of long-term practitioners,the action recognition rate of short-term practitioners is lower,the feature values between different actions are not prominent,and the degree of action discrimination is low;there is obvious difference between the EMG signal feature values of long-term and s h ort-term practitioners The difference can be distinguished by the action recogn i tion rate,which can provide guidance and basis for the improvement of the level of the practitioner.