AbstractThe dynamic weighing signal processing method in high-speed fruits sorting system directly affects the weighing accuracy. This paper describes a basic modeling method for time series, including modeling, order determining, applicability analysis, and prediction. Considering the structural characteristics of weighing system and the limit of PLC sampling speed, a modified auto-regressive (AR) model for stable value prediction of the high-speed weighing signals is presented. Meanwhile, a simulation base on Matlab platform and an implementation example are used to test the method performance. At last, the impacts of environmental changes, sorting speeds, fruits shapes, and sizes on the method performance are discussed. In experimental conditions, this method can be applied to fruits sorting system with sorting speed up to 18 per second for standard fruit (smooth round surface without defects) and an acceptable weighing accuracy can be obtained.