Abstract: Spectrum sensing is one of the key technologies in cognitive radio and energy detection is a typical method of it. According to the noise uncertainty problem of conventional energy detection in spectrum sensing, this paper proposes an adaptive threshold algorithm of energy detection based on noise estimation. Combined with the system requirements, the proposed algorithm adopts constant false-alarm probability detection; it adjusts the threshold parameters adaptively using the real-time noise variance estimation based on the noise statistical model. The simulation results show that the proposed algorithm estimates the noise variance accurately and improves sensing performance obviously.