Accurate recognition and understating of human emotions is an essential skill that can improve the collaboration between humans and machines. In this vein, electroencephalogram (EEG)-based emotion recognition is considered an active research field with challenging issues regarding the analyses of the nonstationary EEG signals and the extraction of salient features that can be used to achieve accurate emotion recognition. In this paper, an EEG-based emotion recognition approach with a novel time-frequency feature extraction technique is presented. In particular, a quadratic time-frequency distribution (QTFD) is employed to construct a high resolution time-frequency representation of the EEG signals and capture the spectral variations of the EEG signals over time. To reduce the dimensionality of the constructed QTFD-based representation, a set of 13 time- and frequency-domain features is extended to the joint time-frequency-domain and employed to quantify the QTFD-based time-frequency representation of the EEG signals. Moreover, to describe different emotion classes, we have utilized the 2D arousal-valence plane to develop four emotion labeling schemes of the EEG signals, such that each emotion labeling scheme defines a set of emotion classes. The extracted time-frequency features are used to construct a set of subject-specific support vector machine classifiers to classify the EEG signals of each subject into the different emotion classes that are defined using each of the four emotion labeling schemes. The performance of the proposed approach is evaluated using a publicly available EEG dataset, namely the DEAPdataset. Moreover, we design three performance evaluation analyses, namely the channel-based analysis, feature-based analysis and neutral class exclusion analysis, to quantify the effects of utilizing different groups of EEG channels that cover various regions in the brain, reducing the dimensionality of the extracted time-frequency features and excluding the EEG signals that correspond to the neutral class, on the capability of the proposed approach to discriminate between different emotion classes. The results reported in the current study demonstrate the efficacy of the proposed QTFD-based approach in recognizing different emotion classes. In particular, the average classification accuracies obtained in differentiating between the various emotion classes defined using each of the four emotion labeling schemes are within the range of $73.8\%$ – $86.2\%$ . Moreover, the emotion classification accuracies achieved by our proposed approach are higher than the results reported in several existing state-of-the-art EEG-based emotion recognition studies.

The aim of this paper is to present a dataset for Wi-Fi-based human activity recognition. The dataset is comprised of five experiments performed by 30 different subjects in three different indoor environments. The experiments performed in the first two environments are of a line-of-sight (LOS) nature, while the experiments performed in the third environment are of a non-line-of-sight (NLOS) nature. Each subject performed 20 trials for each of the experiments which makes the overall number of recorded trials in the dataset equals to 3000 trials (30 subjects × 5 experiments × 20 trials). To record the data, we used the channel state information (CSI) tool [1] to capture the exchanged Wi-Fi packets between a Wi-Fi transmitter and receiver. The utilized transmitter and receiver are retrofitted with the Intel 5300 network interface card which enabled us to capture the CSI values that are contained in the recorded transmissions. Unlike other publicly available human activity datasets, this dataset provides researchers with the ability to test their developed methodologies on both LOS and NLOS environments, in addition to many different variations of human movements, such as walking, falling, turning, and pen pick up from the ground.

This study aims to increase the control's dimensions of the electroencephalography (EEG)-based brain-computer interface (BCI) systems by distinguishing between the motor imagery (MI) tasks associated with fine body-parts of the same hand, such as the wrist and fingers. This in turn can enable individuals who are suffering from transradial amputations to better control prosthetic hands and to perform various dexterous hand tasks. In particular, we present a novel three-stage framework for decoding MI tasks of the same hand. The three stages of the proposed framework are the input, feature extraction, and classification stages. At the input stage, we employ a quadratic time-frequency distribution (QTFD) to analyze the EEG signals in the joint time-frequency domain. The use of a QTFD enables to transform the EEG signals into a set of two-dimensional (2D) time-frequency images (TFIs) that describe the distribution of the energy encapsulated within the EEG signals in terms of the time, frequency, and electrode position. At the feature extraction stage, we design a new convolutional neural network (CNN) architecture that can automatically analyze and extract salient features from the TFIs created at the input stage. Finally, the features obtained at the feature extraction stage are passed to the classification stage to assign each input TFI to one of the eleven MI tasks that are considered in the current study. The performance of our proposed framework is evaluated using EEG signals that were acquired from eighteen able-bodied subjects and four transradial amputated subjects while performing eleven MI tasks within the same hand. The average classification accuracies obtained for the able-bodied and transradial amputated subjects are 73.7% and 72.8%, respectively. Moreover, our proposed framework yields 14.5% and 11.2% improvements over the results obtained for the able-bodied and transradial amputated subjects, respectively, using conventional QTFD-based handcrafted features and a multi-class support vector machine classifier. The results demonstrate the efficacy of the proposed framework to decode the MI tasks associated with the same hand for able-bodied and transradial amputated subjects.

This paper presents a dataset for Wi-Fi-based human-to-human interaction recognition that comprises twelve different interactions performed by 40 different pairs of subjects in an indoor environment. Each pair of subjects performed ten trials of each of the twelve interactions and the total number of trials recorded in our dataset for all the 40 pairs of subjects is 4800 trials (i.e., 40 pairs of subjects × 12 interactions × 10 trials). The publicly available CSI tool [1] is used to record the Wi-Fi signals transmitted from a commercial off-the-shelf access point, namely the Sagemcom 2704 access point, to a desktop computer that is equipped with an Intel 5300 network interface card. The recorded Wi-Fi signals consist of the Received Signal Strength Indicator (RSSI) values and the Channel State Information (CSI) values. Unlike the publicly available Wi-Fi-based human activity datasets, which mainly have focused on activities performed by a single human, our dataset provides a collection of Wi-Fi signals that are recorded for 40 different pairs of subjects while performing twelve two-person interactions. The presented dataset can be exploited to advance Wi-Fi-based human activity recognition in different aspects, such as the use of various machine learning algorithms to recognize different human-to-human interactions.