Sparse-representation-based single-channel source separation, in which an attempt is made to recover each source’s signal using its corresponding sub-dictionary, has attracted many scholars’ attention. The basic premise of this model is that each su