The human behavior recognition research of wearable devices usually extracts the feature values of the sensor data

and then combines the classification algorithm to identify the human behavior.Aiming at the problem of feature extraction and classifier

a fusion model human behavior recognition method（HBRM） is proposed.Firstly

the data collected by the acceleration sensor is converted into a two-dimensional tensor format

and then the convolutional neural network is used to extract the characteristics of the tensor.At the same time

considering that the human behavioral action has strong correlation before and after the time series

long short-term memory network is proposed to identify human behavioral actions.Since convolutional neural networks have good performance in feature extraction

and long-and short-term memory models are good at processing time series problems

the fusion of these two models has good theoretical effect.Experiments are carried out on the WISDM dataset

the result shows that the average recognition rate of the six human behavioral actions reach 96.95 %.