Magnetic Barkhausen noise (MBN) signals typically exhibit anisotropy, which arises from the combined effects of external factors (such as applied stress, magnetic field, and its orientation) and internal factors (such as grain orientation). To systematically investigate the coupling effects of magnetic field direction, easy magnetization axis distribution, and stress on the anisotropy of MBN signals, a force–magnetism synchronous loading system for MBN anisotropy detection was developed in this study. Considering the double easy-axis characteristics within grains of oriented silicon steel, a new anisotropy fitting model for MBN signals was proposed, which demonstrates higher fitting accuracy compared with conventional empirical formulas. Anisotropy detection experiments were carried out on oriented silicon steel specimens subjected to tensile stress along different rolling directions. The results reveal that the influence of tensile stress on MBN anisotropy strongly depends on its angle relative to the easy magnetization axis: when this angle exceeds 45°, stress induces the formation of a double easy-axis system, significantly altering the magnetization behavior. Furthermore, particle swarm optimization was employed to identify and optimize stress-dependent fitting parameters, enabling analysis of their correlation with tensile stress. These findings provide important insights into the application of MBN signals for stress evaluation and magnetic anisotropy characterization.