Constrained non-Negative Matrix Factorization Method for EEG Analysis in Early Detection of Alzheimer Disease

Approximate non-negative matrix factorization (NMF) is an emerging technique with a wide spectrum of potential applications in biomedical data analysis. In this paper, we proposed a new NMF algorithm with temporal smoothness constraint that aims to extract non-negative components that have meaningful physical or physiological interpretations. We propose two constraints and derive new multiplicative learning rules. Specifically, we apply the proposed algorithm, combined with advanced time-frequency analysis and machine learning techniques, to early detection of Alzheimer disease using clinical EEG recordings. Empirical results show promising performance