To extract parameters of the received signal correctly in the digital radio mondiale (DRM) system, robustness mode detection must be accomplished in time domain. Based on the cyclic prefix (CP), this paper presents a robustness mode detection algorithm in the DRM system. Using the system parameters of each robustness mode separately, the linear-weighted correlation and global maximum correlation-value searching algorithm is applied to detect the robustness mode. Simulations show that the proposed correlation operation in the presented algorithm has a better detection performance than the other correlation operations, making the algorithm accurate under flexible parameter choice. The proposed algorithm overcomes the disadvantage of the existing algorithms that can not detect the mode A signal, and the detection accuracy can exceed 99% for each mode at signal-noise-ratio (SNR) no less than 8 dB. When SNR decreases to 0 dB, the accuracy can also achieve 97%.