Applicability of Hyperdimensional Computing to Seizure Detection

Hyperdimensional (HD) computing is a form of brain-inspired which can be applied to numerous classification problems. In past research, it has been shown that seizures detected from electroencephalograms (EEG) with high accuracy using local binary pattern (LBP) encoding. This paper explores applicability HD seizure detection intra-cranial EEG (iEEG) data the Kaggle contest based on both LBP and power spectral density (PSD) features. PSD method, three novel approaches are presented for selected features all These referred as single classifier long hypervector , xmlns:xlink="http://www.w3.org/1999/xlink">multiple classifiers short . To visualize quality test data, xmlns:xlink="http://www.w3.org/1999/xlink">hypervector distance plot introduced plots Hamming distance query hpervectors one class hypervector xmlns:xlink="http://www.w3.org/1999/xlink">vs. other. Simulation results show that: xmlns:xlink="http://www.w3.org/1999/xlink">1) method offers an average 80.9% accuracy, 71.9% sensitivity, 81.4% specificity 76.6% AUC whereas achieve 91.0% 81.8% 92.0% 86.9% AUC. xmlns:xlink="http://www.w3.org/1999/xlink">2) The latency 2.5s 4.5s methods. latency, less than 5s, relevant parameter fast drug delivery, indicating methods able detect in timely manner. performance better xmlns:xlink="http://www.w3.org/1999/xlink">3) It dimensionality reduced 1, 000 bits 10, 000. Futhermore, some features, 100 bits.