Spectrum sensing technique for wideband analog signals has been applied widely in cognitive radio. Based on the idea of time-division, this study was conducted with the following steps. Firstly, sampling time was divided into time slices which have fixed length. Secondly, the signal was multiplied by a bank of periodic waveforms which are modulated by pseudorandom number generators, and the product is low-pass filtered and sampled at sub-Nyquist rate. Finally, the samples were used to evaluate frequency support. Compared with the multi-channel structure, the single-channel structure proposed here is much simpler and has the ability to spectrum sensing with low-rate samples. Numerical simulations show that this algorithm can effectively use samples sampling at sub-Nyquist rate to finish the task of spectrum sensing in scenarios that the spectral support is unknown in advance.