Activity Prediction for Localizing the Events in Imagined Speech EEG Signals

Imagined speech electroencephalogram (EEG) signals are often collected for longer durations than necessary, leading to a difficulty in understanding the generation of EEG during the task as it is likely that most of the data collected is that of resting state. Developing a filter to identify the seg...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:The ... International Winter Conference on Brain-Computer Interface S. 1 - 5
Hauptverfasser: Balasubramanian, Arun, Pandey, Kartik, Veer, Gautam, Samanta, Debasis
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 24.02.2025
Schlagworte:
Accuracy
EEG
Electroencephalography
Feature extraction
feature importance
frequency analysis
Image segmentation
Imagined speech
Information filters
Location awareness
neural decoding
Neuroscience
resting state
Signal processing
Speech processing
Vectors
ISSN:2572-7672
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Imagined speech electroencephalogram (EEG) signals are often collected for longer durations than necessary, leading to a difficulty in understanding the generation of EEG during the task as it is likely that most of the data collected is that of resting state. Developing a filter to identify the segments of the trials with actual information and remove those that contain the EEG from resting states could significantly advance our understanding of EEG signals in neuroscience and biomedical engineering. This work uses derivatives-based features to form the feature vectors used to train the classifiers. Based on feature importance analysis, it has been found that the derivative-based features contributed more to the classification than the traditionally preferred feature, band power in the alpha frequency band. Using the proposed features, the classification performance experienced a significant enhancement, surpassing the results reported in previous studies. The greater accuracy of the classifiers with the proposed features implies that they are effective at filtering out the resting state segments from imagined speech EEG signals.
ISSN:2572-7672
DOI:10.1109/BCI65088.2025.10931413