Detector De Coerência E Modelos Ocultos De Markov Aplicados À Classificação De Tarefas De Imagética Motora

Abstract

Abstract The most investigated stages in Brain-Computer Interface are features extraction and task classification. Thus, this work investigates the application of the coherence detector (Magnitude Squared Coherence -MSC) and the Hidden Markov Models (HMM) in the extraction of features and tasks classification, respectively. Features were extracted from electroencephalogram (EEG) in the Delta band (0.1-2 Hz), Alpha band (8-13 Hz) and Beta band (14-30 Hz) using coherence with 5% and 10% significance level (α). The EEG signals were recorded from three healthy subjects during three events: spontaneous EEG, EEG-based motor task and EEG-based motor imagination. We recorded EEG with electrodes placed according to the international 10-20 (first section) and 10-10 systems (second and third sections). HMM observations were obtained by the coherence calculated with 12 trials and the detected frequency range with higher MSC was adopted as a feature for classifier. The hit rate in classification was 72.5 %, 68 % and 65 % for subjects # 1, # 2 and # 3, respectively, using α=5 %. When we used α=10 %, the rates were 72 %, 57.5 % and 67.5 %. Results shown we can extract features from brain activities related to different events by using coherence detector and that HMM is useful in the classification of imaginary movements.