This letter aims to improve the performance of the multiple antenna sensing by exploiting features beyond eigenvalues and covariance matrices. In particular, a joint phase-eigenvalue difference-based (PED) scheme considering noise uncertainty is proposed. Specifically, PD distributions are independent of noise variance, which means that features robust to noise uncertainty can be extracted from them. The impact of noise uncertainty on eigenvalues is further analyzed, and the extracted robust PD-based features are optimally incorporated with eigenvalues to construct the joint PD-eigenvalue feature and maximize the probability of detection, thus greatly improving the performance of the proposed scheme. Simulation results show that the proposed scheme significantly outperforms the state-of-the-art schemes.