According to the Nyquist sampling theory, it is necessary to have ADCs of capability of GHz sampling speed to sense the spectrumof GHz bandwidth in real time. But this is not possible with the limitations of budget and energy consumption. The BigBand sparse spectrum sensing algorithm, targeting to the sparsity of signals, is capable of sensing 0.9 GHz wide spectrumusing several 50 MHz low speed ADCs. In this paper, the threshold for eliminating out the noise is researched, by using the statistical parameters of noise. A threshold calculating method is proposed, supporting the implements of BigBand algorithm in practices.