Hassan Akbari: In this work, we use tensor decomposition methods in application of fetal ECG extraction from abdominal recordings. We propose a linear source separation algorithm based on a measure of quasi-periodicity. The quasi-periodicity is attained through the use of a constraint on both matrix factorization and tensor decomposition problems. In practice, we form a three dimensional ”tensor” by stacking the observation matrix and rough estimates obtained by both linear and non-linear subspace reconstruction methods.Validation is carried out on a simulated dataset with varying SINRs, where results show us that the proposed method successfully outperforms existing methods in literature.
Reza Sameni: Presentation of his work at Shiraz University in the Signal Processing group.