In this research, we develop an affective computing method based on machine learning for emotion recognition using a wireless protocol and a wearable electroencephalography (EEG) custom-designed device. The system collects EEG signals using an eight-electrode placement on the scalp; two of these electrodes were placed in the frontal lobe, and the other six electrodes were placed in the temporal lobe. We performed experiments on eight subjects while they watched emotive videos. Six entropy measures were employed for extracting suitable features from the EEG signals. Next, we evaluated our proposed models using three popular classifiers: a support vector machine (SVM), multi-layer perceptron (MLP), and one-dimensional convolutional neural network (1D-CNN) for emotion classification; both subject-dependent and subject-independent strategies were used. Our experiment results showed that the highest average accuracies achieved in the subject-dependent and subject-independent cases were 85.81% and 78.52%, respectively; these accuracies were achieved using a combination of the sample entropy measure and 1D-CNN. Moreover, our study investigates the T8 position (above the right ear) in the temporal lobe as the most critical channel among the proposed measurement positions for emotion classification through electrode selection. Our results prove the feasibility and efficiency of our proposed EEG-based affective computing method for emotion recognition in real-world applications.

In this study, we propose an end-to-end emotion recognition system using an earelectroencephalogram (EEG)-based on-chip device that is enabled using the machine-learning model. The system has an integrated device that gathers EEG signals from electrodes positioned behind the ear; it is more practical than the conventional scalp-EEG method. The relative power spectral density (PSD), which is the feature used in this study, is derived using the fast Fourier transform over five frequency bands. Directly on the embedded device, data preprocessing and feature extraction were carried out. Three standard machine learning models, namely, support vector machine (SVM), multilayer perceptron (MLP), and one-dimensional convolutional neural network (1D-CNN), were trained on these rich emotion classification features. The traditional approach, which integrates a model into the application software on a personal computer (PC), is cumbersome and lacks mobility, which makes it challenging to use in real-life applications. Besides, the PC-based system is not sufficiently real-time because of the connection latency from the EEG data acquisition device. To overcome these limitations, we propose a wearable device capable of performing on-chip machine learning and signal processing on the EEG data immediately after the acquisition task for the real-time result. In order to perform on-chip machine learning for the real-time prediction of emotions, 1D-CNN was chosen as a pre-trained model using the relative PSD characteristics as input based on the evaluation of the set results. Additionally, we developed a smartphone application that alerted the user whenever a negative emotional state was identified and displayed the information in real life. Our test results demonstrated the feasibility and practicability of our embedded system for real-time emotion recognition.INDEX TERMS Electroencephalogram (EEG), emotion recognition, tiny machine learning, real-time EEG system, power spectral density (PSD), multilayer perceptron (MLP), support vector machine (SVM), onedimensional convolutional neural network (1D-CNN).

No abstract

In previous investigations two additive 5-step scales (Guttman-scales) for »susceptibility to stimulation« and »reactivity« were developed. Both scales have proved very reliable for the measurement of levels of coma. Compared to usual examinations they entail additional effort. The present paper examines the possible prediction of scale-values, using neurological signs which are easier to obtain than the scale scores. A Bayes procedure was developed and. tested. Results from classification demonstrate that reactivity is highly predictable. Sequential selection of neurological signs, dependent upon their information-content leads to additional simplification of the diagnostic procedure. In contrast, susceptibility to stimulation can only be inadequately predicted from neurological signs.

No abstract