Abstract
•Effectively combines LCADCs, MASA, and adaptive-rate FIR filter for signal denoising.•Reduces data size using adaptive-rate techniques, features extraction, and selection.•Proposes ensemble of sub-problems classification to separate closely related classes.
Epilepsy is a disorder of the brain characterized by seizures and requires constant monitoring particularly in serious patients. Electroencephalogram (EEG) signals are frequently used in epilepsy diagnosis and monitoring. A new paradigm of battery packed wearable gadgets has recently gained popularity which constantly monitor a patient’s signals. These gadgets acquire the data and transmit it to the cloud for further processing. Power consumption due to data transmission is a major issue in these devices. Moreover, in a constant monitoring environment, the number of classes to be identified are usually higher and overlapping. Existing techniques either require the entire data to be transmitted, such as in deep learning, or suffer from reduced accuracy. In this context, we propose a new framework for EEG based epilepsy detection which requires a low data transmission while maintaining high accuracy for multiclass classification. At the device-end, we use a preprocessing mechanism that uses adaptive rate sampling, modified activity selection, filtering, and wavelet decomposition to extract only a handful of highly discriminatory features to be transmitted instead of the entire EEG waveform. For multiclass classification, we propose a novel ensemble of sub-problems-based classification paradigm to achieve high accuracy using the reduced data. Our proposed solution shows many-fold increase in computational gains and an accuracy of 100% and 99.38% on the 2-class problem when tested on the popular University of Bonn and CHB-MIT datasets, respectively. An accuracy of 99.6% on 3-class, 96% on 4-class, and 92% on 5-class problems is obtained for the University of Bonn dataset.