In order to solve the problem of low weight detection precision of belt type weight detection equipment at high speed. Firstly, the model of weight detection system was modeled, and then the checkweigher in static, no-load operation, the normal work of the state of the system was checked to analyze time frequency signals.The source of interference and its frequency were found out, then an analysis of its causes and changes were analyzed, determining the index filtering. According to the index, the FIR digital filter was designed to denoise the data. After filtering, two methods were adopted as follows. The weight estimation of asymmetric trimmed mean (Trimmed-mean mass estimator, referred to as TME) was investigated. After taking the last N data and sorting by size, this sequence was cut in an asymmetrical way for average. Moreover, the arithmetic average weight estimation (Averagemassestimation, referred to as AME) used to intercept the last N data, and the average was adopted. The results showed that the weight error of TME was less than 0.3%, and the average error is 0.155%, with the average error of AME 0.758%. It was concluded that the FIR filtering could fast denoise the wave, and the TME weight estimation had higher detection accuracy.