In recent years,some scholars introduced dynamic Allan variance into the random signal processing of inertial components to implement real-time tracking of random signal error.However,since the dynamic Allan variance uses the function with fixed window length for intercepting random signals,it can not guarantee both the signal dynamic tracking performance and variance confidence level.To solve the problem,this paper proposes a dynamic Allan variance arithmetic based on entropy of information.First,the entropy of information in the window is used for representing the non-stationarity of the dynamic gyro signal.Then,it can adaptively select the appropriate window length for the next moment according to the entropy of random signal,and calculate the Allan variance of the signal.Simulation result proves that the algorithm can effectively improve the real-time tracking capabilities of the gyro dynamic signal.