Toward asynchronous EEG-based BCI: Detecting imagined words segments in continuous EEG signals

Abstract

An asynchronous Brain-Computer Interface (BCI) based on imagined speech is a tool that allows to control an external device or to emit a message at the moment the user desires to by decoding EEG signals of imagined speech. In order to correctly implement these types of BCI, we must be able to detect from a continuous signal, when the subject starts to imagine words. In this work, five methods of feature extraction based on wavelet decomposition, empirical mode decomposition, frequency energies, fractal dimension and chaos theory features are presented to solve the task of detecting imagined words segments from continuous EEG signals as a preliminary study for a latter implementation of an asynchronous BCI based on imagined speech. These methods are tested in three datasets using four different classifiers and the higher F 1 scores obtained are 0.73, 0.79, and 0.68 for each dataset, respectively. This results are promising to build a system that automatizes the segmentation of imagined words segments for latter classification.

AU-NN: ANFIS Unit Neural Network

“…This is a two class classification problem with IL. To realize this task, the method of signal segmentation used in [16] was applied to obtain the exact same IWS and ISS instances. From the three datasets used in that work, the first dataset, the one recorded in [17] was selected, this is because it is the only one from the three datasets that was recorded in sessions, this is a characteristic that we need in order to train and test the model in an IL approach.…”

“…From the three datasets used in that work, the first dataset, the one recorded in [17] was selected, this is because it is the only one from the three datasets that was recorded in sessions, this is a characteristic that we need in order to train and test the model in an IL approach. From all the feature sets used in [16] the one based on Instant Wavelet Energies (IWE) was selected, this is because for the given dataset this feature set was the one that reported the higher results.…”

“…4. For the train sets, from each trial in each session, segments are extracted with a moving window of size 0.5 seconds and an overlap of 0.1 seconds, and features are extracted as in [16]. With this, each trial is turned into a set of instance vectors of IWS's and ISS's.…”

“…(i) The classification of imagined words in EEG, and (ii) The detection of Imagined Word Segments (IWS) vs. Idle State Segments (ISS) [16] in an incremental learning way. In addition, a process of classification is described, where instead of taking the whole instance and taking it as one example, each instance is decomposed into a set of smaller instances, and the classification is done by a majority vote over all the predictions of the set.…”

DAL: Feature Learning from Overt Speech to Decode Imagined Speech-based EEG Signals with Convolutional Autoencoder

“…They conducted the classification using the shrinkage regularized linear discriminant analysis and the random forest (RF) which are general machine learning algorithms. Hernández-Del-Toro et al [20] decoded imagined speech-based EEG signals using five feature extraction methods to solve the task of detecting imagined words segments from continuous EEG signals. They tested in three datasets using four machine learning algorithms (RF, k-nearest neighbors, SVM, and logistic regression).…”