The human gait as a source of biometric data has improved identification at father distance, however it only full body gait data has been explored with deep learning models, which is resource demanding not always available due obstacles as in environment of deployment. In this research, we explored the use of upper gait analysis for identification with deep learning model convolutional recurrent neural network- Long Short Term Memory CRNN-LSTM and evaluate the reliability of using half gait against full gait, primary dataset was collected from 26 subject 12 females and 14 males. The result returns better accuracy with upper half gait than full body gait, hence, lower computation demand.