Emotion Detection from Electroencephalography Signals Using String Grammar K-Nearest Neighbors
Kampanat Sutijirapan, Sansanee Auephanwiriyakul, Nipon Theera‐Umpon · 2024
Electroencephalography (EEG) signals-based motion detection has become more popular recently. There are several methods proposed to ease this application. However, in this paper, we apply a different approach, i.e., we explore the syntactic approach in detecting emotion from the EEG signals. In particular, we develop a translation method to translate an EEG signal into a string. Then we utilize the string grammar$k{-}$NN method to find a suitable class for each test signal. However, since it is a string calculation, we have to use the Levenshtein distance instead of a regular distance calculation. From the validation test results, we found that the maximum accuracy rate is 96.95% in the$4^{\text{th}}$validation with$k=3$. For the test results with the same model, we found that the accuracy rate is around 87 - 89%. This shows that this syntactic pattern recognition can provide comparable results with existing methods.