Novel Adaptive Kalman Filter with Fuzzy Neural Network for Trajectory Estimation System

The objective of this paper presents a novel trajectory estimation (TE) system for mitigating the measurement noise and the undulation for the implementation of the touch interface. The methodology is based on an adaptive Kalman filter with a fuzzy neural network (FNN). The FNN is implemented as an artificial intelligence decision maker to regulate the smoothness of the output estimation. In real time, the FNN adaptively configures the filtering performance of the Kalman filter by analyzing the trajectory using the Fourier spectrum and the kinematic data. The result of this research is demonstrated on the touch interface that is implemented using the IR camera sensing method, and the proposed TE system is embedded to a coordinate processing module that converts the raw touch coordinates into the USB human interface device multi-touch protocol. A 70-inch projector screen is set up for the experiment, and the touch trajectories are tested with various magnitudes of measurement noise. The significance of the proposed TE system is demonstrated by showing the tracking delay and the distortion of the filtered trajectory are adaptively reduced. As the magnitude of the measurement noise increases, the proposed TE system detects the unwanted high-frequency component and decreases the filtering gain to stabilize the smoothness of the output trajectory. In conclusion, when compared to the recent researches, the proposed TE system shows lower tracking error and tolerates high magnitude of randomly appeared noise on the touch interface.