In this paper, based on the information theoretic learning theory, a novel non-Gaussian filter is proposed in the frame of Kalman filter under Huber correntropy loss function, which is short for HCKF (Huber correntropy Kalman filter). Firstly, a novel Huber correntropy loss function is defined. Secondly, HCKF is designed to solve the Huber correntropy loss function by using a fixed-point iterative algorithm and its superior performance is shown by compared Huber correntropy loss function with the L2 loss function, the Huber loss function, and the Gaussian kernel function. The validity of the HCKF is verified by comparing it with three representative filtering methods in the final simulation.