Windowing compensation in Fourier based Surrogate Analysis and application to EEG signal classification


Téléchargements par mois depuis la dernière année

Plus de statistiques...

Caza-Szoka, M. et Massicotte, D. (2022). Windowing compensation in Fourier based Surrogate Analysis and application to EEG signal classification. IEEE Transactions on Instrumentation and Measurement, 71 . pp. 1-11. ISSN 1557-9662 DOI 10.1109/TIM.2022.3149325

[thumbnail of MASSICOTTE_D_1_POST.pdf]
Télécharger (720kB) | Prévisualisation


This paper shows how adding a second step of windowing after each phase randomization can reduce the False Rejection Rate in Fourier based Surrogate Analysis. Windowing techniques reduce the discontinuities at the boundaries of the periodically extended data sequence in Fourier Series. However, they add a time domain non-stationarity which affects the Surrogate Analysis. This effect is particularly problematic for short low-pass signals. Applying the same window to the surrogate data allows having the same non-stationarity. The method is tested on order 1 autoregressive process null hypothesis by Monte-Carlo simulations. Previous methods were not able to yield good performances for left-sided and right-sided tests at the same time, even less with bilateral tests. It is shown that the new method is conservative for unilateral tests as well as bilateral tests. In order to show that the proposed windowing method can be useful in real context, in this extended paper, it was applied for an EEG diagnostic problem. A dataset comprising the EEG measurements of 15 subject distributed in three groups: attention-deficit disorder primarily hyperactive-impulsive (ADHD), attention-deficit disorder primarily inattentive (ADD); and anxiety with attentional fragility (ANX) was used. Both statistical and machine learning (Naïve Bayesian) approaches were considered. The Mean Short Windowed SA (MSWSA) was used as a signal feature and its performances was studied with respect to the windowing systems. The main findings were that (i) the MSWSA feature has less variability for ADD than for ADHD or ANX, (ii) the proposed windowing method reduces bias and non-normality of the SA feature, (iii) with the proposed method and a naïve Bayesian classifier, a 93% success rate of discriminating ADD from ADHD and ANX was achieved with leave-one-out cross-validation, and (iv) the new feature could not have yielded interesting results without the proposed windowing system.

Type de document: Article
Mots-clés libres: ADD ADHD EEG Fractal Dimension Nonlinear analysis Nonlinear dynamics Surrogate data Windowing techniques
Date de dépôt: 09 mai 2022 20:05
Dernière modification: 09 mai 2022 20:05
Version du document déposé: Post-print (version corrigée et acceptée)

Actions (administrateurs uniquement)

Éditer la notice Éditer la notice