1. Ultra-fast Epileptic seizure detection using EMD based on multichannel electroencephalogram
- Author
-
Yan-Yu Lam, Ming-Jang Chiu, Jeng-Wei Lin, Wei Chen, Hsiao-Ya Sung, Chia-Ping Shen, and Feipei Lai
- Subjects
medicine.diagnostic_test ,Computer science ,business.industry ,Speech recognition ,Gaussian ,Linear model ,Pattern recognition ,Probability density function ,Electroencephalography ,medicine.disease ,Hilbert–Huang transform ,Epilepsy ,symbols.namesake ,medicine ,symbols ,Spike (software development) ,Epileptic seizure ,Artificial intelligence ,medicine.symptom ,business - Abstract
We present a system to detect seizure and spike in Epilepsy Electroencephalogram (EEG) analysis and characterize different epilepsy EEG types. After extracting features from three EEG types, Normal, Seizure and Spike, with Empirical Mode Decomposition (EMD), we do Analysis of variance (ANOVA) to classify conspicuous features and low-resolution features, and build Gaussian distributions of conspicuous features for probability density function (PDF) to do classification. Using EMD, the recognition rate improved from 70% to 90%. With ANOVA, the recognition rate can reach 99%. The linear model accelerates the system from 2 hours to 90 seconds compare to the previous approach.
- Published
- 2013
- Full Text
- View/download PDF