Tensor Based Singular Spectrum Analysis For Nonstationary Source Separation
Samaneh Kouchaki, Saeid Sanei

Tensor based singular spectrum analysis (SSA) has been introduced as an extension of traditional singular value decomposition (SVD) based SSA. In the SSA decomposition stage PARAFAC tensor factorization has been employed. Using tensor factorization methods enable SSA to perform much better in nonstationary and underdetermined cases. The results of applying the proposed method to both synthetic and real data show that this system outperforms the original SSA, when used for single channel data decomposition in nonstationary and underdetermined source separation.