Abstract: To address the issue of bias instability in hemispherical resonator gyroscopes (HRGs) under full-angle mode caused by temperature drift and standing wave drift, a compensation scheme is proposed based on BP neural networks. Firstly the influence mechanism of temperature and standing wave angle is investigated on bias. A BP neural network model incorporating both temperature and standing wave angle as variables is constructed. Three sets of data are used in the experimental design for comparative verification, and the effectiveness of least-squares polynomial compensation and BP neural network compensation is compared through measured data. The results show that both methods can effectively improve the bias stability of HRGs, with the BP neural network compensation performing more optimally. The three experimental datasets demonstrate that the bias stability of the compensated data using the BP network is improved by over 90%, verifying the advantage of this method in suppressing multi-source error coupling. The research results provide a theoretical basis and engineering reference for enhancing the output accuracy of hemispherical resonator gyroscopes.