Automatic seizure detection using a highly adaptive directional time–frequency distribution

Nabeel Ali Khan; Ali Akbar Pouyan; Mokhtar Mohammadi

Abstract:

Electroencephalogram (EEG) signals can be used by a proficient neurologist to detect the presence of seizure activity inside the brain. Automated detection of seizures in EEG signals has clinical importance given that manual round-the-clock monitoring of EEG signals is impossible. A patient-independent algorithm for seizure detection is developed using features extracted from high-resolution time–frequency distributions (TFDs). In order to achieve good classification performance, a modified highly adaptive time–frequency distribution (HADTFD) is defined. The modified-HADTFD is used to obtain a clear and cross-term free time–frequency representation of EEG signals. This is followed by the extraction of features and training of a linear classifier. The proposed approach based on modified-HADTFD achieves the classification accuracy of 98.56% by using only three time–frequency features, which is 37% more than the accuracy achieved with other TFDs

research from:

• Foundation University - Faculty of Engineering and IT

Year:

2018

Type of Publication:

Article

Journal:

Multidimensional Systems and Signal Processing

Volume:

29

Number:

4

Pages:

1661-1678

Month:

10

DOI:

10.1007/s11045-017-0522-8