Abstract: To further improve the detection performance of power spectrum-based spectrum sensing algorithms, we propose estimated noise variance using the geometric mean of the power spectrum and the arithmetic mean of sorted minimum values. Based on these two estimates, along with the maximum and minimum values of the power spectrum, a detection statistic was constructed. A fused spectrum sensing algorithm based on the power spectrum was proposed, and its false alarm probability and theoretical threshold were derived. Simulations under additive white Gaussian noise and Rayleigh fading channels demonstrate superior detection performance compared with existing methods, enhancing frequency offset resistance while maintaining strong noise power uncertainty robustness. Practical signal detection results validate the algorithm's effectiveness.