Abstract: The cognitive frequency hopping technology is considered to be an effective approach to eliminate dilemma in frequency use of traditional frequency hopping systems. To fulfill technical requirements of ultrawideband, multifrequencyslot and real time spectrum sensing, a novel spectrum sensing algorithm based on bidirectional search of normalized powerspectrum (BSNP) used in cognitive frequency hopping is proposed. The BSNP takes the normalized spectrum of frequency slots as the detection statistics, and finds out all occupied frequency slots in frequency hopping bandwidth by executing forward and reverse searches in sequence. The BSNP algorithm makes use of asymptotic normality and independence of Fourier transform to derive the mathematical expressions for the decision threshold and the probabilities of false alarm of single decision. Theoretical analysis and simulation results show that the proposed algorithm can accurately identify occupied frequency slots. Compared to the conventional spectral estimation spectrum sensing algorithm, the BSNP algorithm can effectively overcome the noise uncertainty problem in spectrum sensing.