A Hybrid Feature Pool-Based Emotional Stress State Detection Algorithm Using EEG Signals

Abstract

Human stress analysis using electroencephalogram (EEG) signals requires a detailed and domain-specific information pool to develop an effective machine learning model. In this study, a multi-domain hybrid feature pool is designed to identify most of the important information from the signal. The hybrid feature pool contains features from two types of analysis: (a) statistical parametric analysis from the time domain, and (b) wavelet-based bandwidth specific feature analysis from the time-frequency domain. Then, a wrapper-based feature selector, Boruta, is applied for ranking all the relevant features from that feature pool instead of considering only the non-redundant features. Finally, the k-nearest neighbor (k-NN) algorithm is used for final classification. The proposed model yields an overall accuracy of 73.38% for the total considered dataset. To validate the performance of the proposed model and highlight the necessity of designing a hybrid feature pool, the model was compared to non-linear dimensionality reduction techniques, as well as those without feature ranking.

Detection of Negative Stress through Spectral Features of Electroencephalographic Recordings and a Convolutional Neural Network

“…To this respect, Table 4 shows information from recent studies dealing with stress recognition using EEG spectral features and machine learning classifiers. As can be observed, the classification results obtained by these works presented accuracy values between 57 and 80% [15][16][17][18]52], which are significantly lower than those obtained in the present study by all pre-trained 2-D and 3-D AlexNet-based networks.…”

“…It is important to remark that these studies must be compared with caution, because most of them used different methodologies and experimental protocols, where substantial changes in the number of participants, EEG electrodes, and emotion elicitation ways were found. Nonetheless, two studies included in Table 4 , i.e., works [ 15 , 52 ], used the same public database as in the present work, and thus a direct and fair comparison could be established. In this case, the use of pre-trained 2-D and 3-D CNN-based classification models has reported an improvement of around 10–15% with respect to the accuracy obtained by common machine learning classifiers in [ 15 , 52 ].…”

Health State Classification of a Spherical Tank Using a Hybrid Bag of Features and K-Nearest Neighbor

“…From the hit vector, the final setoff features are acquired. The whole of the techniques from the start till this part are echoed until the importance is designated for all the provided set of feature attributes [11,14]. The steps of this algorithm are demonstrated in Figure 4.…”

“…This algorithm works on three main principles: (1) determine the space among the neighbors, (2) find the k closest neighbors, and (3) vote for labels. Figure 5 illustrates the details of the k-NN algorithm [11,13]. In this study, the considered dataset is allocated with a 70:30-train: test ratio.…”

“…However, to select the most distinctive feature set from the proposed HBoF, it is logical to determine the underlying cause for multiple health conditions. Unlike the popular feature selector algorithms (i.e., GA), Boruta can select all the appropriate features from the feature set instead of selecting only the non-redundant ones [11]. Therefore, in this study, a wrapper-based feature selector algorithm-Boruta is utilized as the feature selector.…”