Recognition of EEG signals of dyslexic children using long short-term memory

Abstract

An investigation into an advanced method of diagnosing dyslexia in children is required to overcome the limitations of the conventional technique. Electroencephalograms (EEG) can divulge brain activities and hence detect dyslexia through proper digital signal processing combined with Long-Short Term Memory (LSTM). This paper describes the recognition of EEG signals of dyslexic and normal children using LSTM. The EEG signals were acquired from the subjects during writing, analysed and fed into the LSTM classifier without passing through the extraction process. Using the optimal parameters obtained via the heuristic approach, the LSTM was able to distinguish the EEG signals of dyslexic and normal children with an average accuracy of 99.3% for training and 86.2% for testing.