Deep Learning of Multifractal Attributes from Motor Imagery Induced EEG

Electroencephalogram (EEG) is an effective metric to monitor or measure human brain activities. Another advantage for EEG utilization is noninvasive, and is not harmful to subjects. However, this leads to two drawbacks: low signal-to-noise ratio and non-stationarity. In order to extract useful features contained in the EEG, multifractal attributes were explored in this paper. A few attributes were utilized to analyze the EEG recorded during motor imagery tasks. Then, we built a deep learning model based on denoising autoencoder to recognize different motor imagery tasks. From the results, we can find that 1) Motor imagery induced EEG is of multifractal attributes, 2) multifractal spectrum D(h) and the statistics cp based on cumulants can reflect difference between different motor imagery tasks, so they can be adopted as features for classification. 3) A deep network with initialization by denoising autoencoder is suitable to learn multifractal attributes extracted from EEG. The classification accuracies demonstrated that the proposed method is feasible.

[1]  Liqing Zhang,et al.  A competitive brain computer interface: Multi-person car racing system , 2013, 2013 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC).

[2]  Nacim Betrouni,et al.  Fractal and multifractal analysis: A review , 2009, Medical Image Anal..

[3]  E Oczeretko,et al.  Fractal geometric analysis of lung cancer angiogenic patterns. , 2001, Folia histochemica et cytobiologica.

[4]  D. Ghosh,et al.  Multifractal parameters as an indication of different physiological and pathological states of the human brain , 2014 .

[5]  Patrice Abry,et al.  Multifractality Tests Using Bootstrapped Wavelet Leaders , 2007, IEEE Transactions on Signal Processing.

[6]  G. Pfurtscheller,et al.  ‘Thought’ – control of functional electrical stimulation to restore hand grasp in a patient with tetraplegia , 2003, Neuroscience Letters.

[7]  Tzyy-Ping Jung,et al.  Combining ERPs and EEG spectral features for decoding intended movement direction , 2012, 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society.

[8]  B. Mandelbrot.  How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension , 1967, Science.

[9]  Liqing Zhang,et al.  Spectral Power Estimation for Unevenly Spaced Motor Imagery Data , 2013, ICONIP.

[10]  Khan M Iftekharuddin,et al.  Statistical analysis of fractal-based brain tumor detection algorithms. , 2005, Magnetic resonance imaging.

[11]  Pascal Vincent,et al.  Stacked Denoising Autoencoders: Learning Useful Representations in a Deep Network with a Local Denoising Criterion , 2010, J. Mach. Learn. Res..

[12]  R. Cortvrindt,et al.  Confocal microscopy: principles and applications to the field of reproductive biology. , 2001, Folia histochemica et cytobiologica.

[13]  Jie Li,et al.  Design of assistive Wheelchair System directly Steered by Human Thoughts , 2013, Int. J. Neural Syst..

[14]  P. Abry,et al.  Bootstrap for Empirical Multifractal Analysis , 2007, IEEE Signal Processing Magazine.

[15]  Anatole Lécuyer,et al.  Exploring two novel features for EEG-based brain-computer interfaces: Multifractal cumulants and predictive complexity , 2010, Neurocomputing.

[16]  Liqing Zhang,et al.  Active training paradigm for motor imagery BCI , 2012, Experimental Brain Research.