In order to collect the data of human movements on foot and analyze

a smart shoe system is designed and implemented

which can collect plantar pressure

acceleration and angular velocity data. Besides the shoe body itself

the smart shoe system consists of pressure insole module

wireless transmission module and movement data receiving software. After movements data are collected

the data are transferred to receiving software through Bluetooth

then displayed on screen and stored. Through multi-layer long short-term memory（ LSTM） network

the three kinds of actions can be identified

include sitting

standing and walking. The final experiment verifies that the LSTM recognition model is accuracy

the data collected through smart shoe is effective

and the smart shoes can be used in research of human action recognition.