Wearable Robot Sensor Signal Prediction Algorithm Analysis and Study based on Particle Filtering

: In this paper, a simple platform for exoskeleton booster is designed and built. The sensing device uses static torque sensor for real-time signal acquisition and feedback. In order to remove signal impurities, this paper designs simple and effective impurity filtering circuits and algorithms. The filtered sensor signal was analyzed by unit root test to determine its stability, and the initial model was determined by ACF and PACF order method. According to the characteristics of the sensor signals of this system, the MWDAR model is verified by the model order and the short-term sliding window size. Using the existing MWDAR model, the prediction value is low. Because of the low prediction accuracy, this paper proposes to introduce particle filter algorithm to optimize, and design a new sensor signal time series prediction algorithm, and simulate it through MATLAB software. Verify the effectiveness of the designed algorithm. Due to the characteristics of the force sensor, the dynamic response frequency is significantly lower than the human neural response frequency. At the same time, after the particle optimization algorithm is used, the calculation amount is increased, which makes the prediction delay. In response to these problems, this paper uses frequency doubling technology, which can double the dynamic response frequency of the booster sensing system, thus providing a basis for accurate, real-time signal prediction.