. . . "S" . . "978-80-01-05506-9" . "This contribution is focused on optimizing the parameters of a classification system based on HMM used for movement-type classification from EEG recordings. A wide variety of settings of N-times repeated K-fold cross-validation is explored in order to find the optimum setting that balances the quality and stability of results against computational cost. 10-times repeated 5-fold cross-validation was chosen. Careful selection of frequency bands for feature selection in the range of 5-45 Hz was performed. Using feature optimization average classification score improvement of 3.5% was achieved. For the implementation of the experiments a parallel extension of the current EEG toolbox was created, this speeds up the classification up to 4.4 times."@en . "Dobi\u00E1\u0161, Martin" . "Optimization of movement EEG klassification parametrs"@en . "Optimalizace parametr\u016F klasifikace pohybov\u00E9ho EEG" . . . "RIV/68407700:21230/14:00219203" . "Praha" . "Praha" . "Optimalizace parametr\u016F klasifikace pohybov\u00E9ho EEG"@cs . . . "Optimalizace parametr\u016F klasifikace pohybov\u00E9ho EEG" . . . . "Optimalizace parametr\u016F klasifikace pohybov\u00E9ho EEG"@cs . "Tento p\u0159\u00EDsp\u011Bvek je zam\u011B\u0159en na optimalizaci parametr\u016F klasifika\u010Dn\u00EDho syst\u00E9mu na b\u00E1zi HMM slou\u017E\u00EDc\u00EDho k rozpozn\u00E1n\u00ED typ\u016F pohyb\u016F z nahr\u00E1vek pohybov\u00E9ho EEG. Je prozkoum\u00E1n \u0161irok\u00FD v\u00FDb\u011Br nastaven\u00ED N-kr\u00E1t opakovan\u00E9 K-fold k\u0159\u00ED\u017Eov\u00E9 validace, aby bylo nalezeno optim\u00E1ln\u00ED nastaven\u00ED, kter\u00E9 dob\u0159e vyva\u017Euje kvalitu a stabilitu v\u00FDsledk\u016F proti v\u00FDpo\u010Detn\u00ED n\u00E1ro\u010Dnosti. Je vybr\u00E1na 10-kr\u00E1t opakovan\u00E1 5-fold k\u0159\u00ED\u017Eov\u00E1 validace. Je proveden d\u016Fkladn\u00FD v\u00FDb\u011Br frekven\u010Dn\u00EDho p\u00E1sma pro v\u00FDb\u011Br p\u0159\u00EDznak\u016F pro klasifikaci v rozsahu 5 a\u017E 45 Hz. Pomoc\u00ED optimalizace p\u0159\u00EDznak\u016F je na datab\u00E1zi pohyb\u016F ramena a ukazov\u00E1\u010Dku dosa\u017Eeno pr\u016Fm\u011Brn\u00E9ho vylep\u0161en\u00ED klasifika\u010Dn\u00EDho sk\u00F3re o 3,5%. V r\u00E1mci implementace experiment\u016F bylo vytvo\u0159eno paraleln\u00ED roz\u0161\u00ED\u0159en\u00ED st\u00E1vaj\u00EDc\u00EDho EEG toolboxu, kter\u00E9 zrychluje klasifikaci a\u017E 4,4 kr\u00E1t." . "RIV/68407700:21230/14:00219203!RIV15-MSM-21230___" . . "IV. Letn\u00ED doktorandsk\u00E9 dny 2014" . "5"^^ . . "1"^^ . "Optimization of movement EEG klassification parametrs"@en . "34850" . . "2014-05-29+02:00"^^ . "[A5E2B3A87205]" . . "\u010Cesk\u00E9 vysok\u00E9 u\u010Den\u00ED technick\u00E9 v Praze. Fakulta elektrotechnick\u00E1. Katedra teorie obvod\u016F" . "1"^^ . . "21230" . "BCI; EEG; HMM; Movement EEG classification"@en . "Tento p\u0159\u00EDsp\u011Bvek je zam\u011B\u0159en na optimalizaci parametr\u016F klasifika\u010Dn\u00EDho syst\u00E9mu na b\u00E1zi HMM slou\u017E\u00EDc\u00EDho k rozpozn\u00E1n\u00ED typ\u016F pohyb\u016F z nahr\u00E1vek pohybov\u00E9ho EEG. Je prozkoum\u00E1n \u0161irok\u00FD v\u00FDb\u011Br nastaven\u00ED N-kr\u00E1t opakovan\u00E9 K-fold k\u0159\u00ED\u017Eov\u00E9 validace, aby bylo nalezeno optim\u00E1ln\u00ED nastaven\u00ED, kter\u00E9 dob\u0159e vyva\u017Euje kvalitu a stabilitu v\u00FDsledk\u016F proti v\u00FDpo\u010Detn\u00ED n\u00E1ro\u010Dnosti. Je vybr\u00E1na 10-kr\u00E1t opakovan\u00E1 5-fold k\u0159\u00ED\u017Eov\u00E1 validace. Je proveden d\u016Fkladn\u00FD v\u00FDb\u011Br frekven\u010Dn\u00EDho p\u00E1sma pro v\u00FDb\u011Br p\u0159\u00EDznak\u016F pro klasifikaci v rozsahu 5 a\u017E 45 Hz. Pomoc\u00ED optimalizace p\u0159\u00EDznak\u016F je na datab\u00E1zi pohyb\u016F ramena a ukazov\u00E1\u010Dku dosa\u017Eeno pr\u016Fm\u011Brn\u00E9ho vylep\u0161en\u00ED klasifika\u010Dn\u00EDho sk\u00F3re o 3,5%. V r\u00E1mci implementace experiment\u016F bylo vytvo\u0159eno paraleln\u00ED roz\u0161\u00ED\u0159en\u00ED st\u00E1vaj\u00EDc\u00EDho EEG toolboxu, kter\u00E9 zrychluje klasifikaci a\u017E 4,4 kr\u00E1t."@cs .