Using steady state visual evoked potential (SSVEP) brain-machine interface (BMI) system for a long time will cause strong visual fatigue. Reducing the number of redundant flicker stimuli can effectively reduce the area of visual stimuli on the screen, thus improving the user experience and reducing visual fatigue when using the SSVEP-BCIs. The steady state visual peripheral visual evoked potential (SSPVEP) paradigm removes redundant stimuli. At present, there are few signal processing and detection algorithms designed according to the characteristics of EEG signals under SSPVEP paradigm. In this paper, a hierarchical detection method is proposed to improve the performance of EEG signal detection under SSPVEP paradigm. The proposed method makes full use of the applicable characteristics of distance similarity and correlation coefficient similarity in signal detection. This method can effectively improve the detection performance of EEG signals under SSPVEP paradigm when the sampling time is set from 0.2 s to 4.6 s. The detection accuracy of this algorithm is 72.79% when the sampling time is 4.6s, while that of the previous algorithm is 71.54%.