The present invention relates to a compressive sensing-based signal processing method and apparatus. The method includes: determining distribution, of signal components of an input signal on which frequency mixing has been performed, in a baseband spectrum according to spectral distribution of the input signal; dividing the baseband spectrum according to different signal aliasing patterns formed by the distribution of the signal components, to form multiple frequency bands; determining lowest sampling frequency of the input signal according to the signal aliasing patterns in the frequency bands, and performing sampling on the input signal at sampling frequency greater than the lowest sampling frequency, to obtain a sampled signal; and separately restoring a corresponding signal component in each frequency band of the multiple frequency bands according to the sampled signal, and splicing the signal components restored in the frequency bands, to complete restoration of the input signal.