Stockwell transform and semi-supervised feature selection from deep features for classification of BCI signals

Abstract

Over the past few years, the processing of motor imagery (MI) electroencephalography (EEG) signals has been attracted for developing brain-computer interface (BCI) applications, since feature extraction and classification of these signals are extremely difficult due to the inherent complexity and tendency to artifact properties of them. The BCI systems can provide a direct interaction pathway/channel between the brain and a peripheral device, hence the MI EEG-based BCI systems seem crucial to control external devices for patients suffering from motor disabilities. The current study presents a semi-supervised model based on three-stage feature extraction and machine learning algorithms for MI EEG signal classification in order to improve the classification accuracy with smaller number of deep features for distinguishing right- and left-hand MI tasks. Stockwell transform is employed at the first phase of the proposed feature extraction method to generate two-dimensional time–frequency maps (TFMs) from one-dimensional EEG signals. Next, the convolutional neural network (CNN) is applied to find deep feature sets from TFMs. Then, the semi-supervised discriminant analysis (SDA) is utilized to minimize the number of descriptors. Finally, the performance of five classifiers, including support vector machine, discriminant analysis, k-nearest neighbor, decision tree, random forest, and the fusion of them are compared. The hyperparameters of SDA and mentioned classifiers are optimized by Bayesian optimization to maximize the accuracy. The presented model is validated using BCI competition II dataset III and BCI competition IV dataset 2b. The performance metrics of the proposed method indicate its efficiency for classifying MI EEG signals.

Designing Dispatching Rules via Novel Genetic Programming with Feature Selection in Dynamic Job-Shop Scheduling

“…A threshold value ( ) is then computed for penalizing the candidates in subsequent steps (line 4). After the previous initial steps, the survivors are chosen from the candidate set to create the new population through − 1 iterations (see lines [5][6][7][8][9][10][11][12][13][14][15]. The algorithm categorizes the candidates into the penalized set ( ) and the non-penalized set ( ) at each iteration (line 6).…”

“…Previous analysis indicates that the NGP-based algorithm has a much larger advantage over the SGP-based algorithm in terms of regular structure. Thus, Equations ( 13) and (14) show the two example rules obtained by NGP and NGP-FS for further comparison in the MWT objective, respectively. Note that rule 2 has a smaller size than rule 1.…”

“…Feature selection in machine learning provides a good idea for solving this challenging task. It has been proven to be an effective way when dealing with classification [13,14], clustering [15], and regression tasks [16]. Although GP can identify relevant features and use them to evolve the best GP trees simultaneously in the adaptive evolutionary process, there are still some irrelevant or redundant features.…”

Designing Dispatching Rules via Novel Genetic Programming with Feature Selection in Dynamic Job-Shop Scheduling

“…Feature selection in machine learning provides a good idea for solving this challenging task. It has been proven to be an effective way when dealing with classification 14,15 , clustering 16 and regression tasks 17 . Although GP can identify relevant features and use them to evolve the best GP trees simultaneously in the adaptive evolutionary process, there are still some irrelevant or redundant features.…”

“…A threshold value (𝐷) is then computed for penalizing the candidates in subsequent steps (line 4). After the previous initial steps, the survivors are chosen from the candidate set to create the new population through 𝑛 − 1 iterations (see lines[5][6][7][8][9][10][11][12][13][14][15].The algorithm categorizes the candidates into penalized set (𝐶 𝑝 ) and non-penalized set (𝐶 𝑛𝑝 ) at each iteration (line 6). To be specific, any candidate with a distance to the nearest survivor is below the threshold 𝐷, is classified as a penalty candidate or else as a non-penalized candidate.…”