We present the overall design and the implementation of the CLAS dataset, a multimodal resource which was purposely developed in support of research and technology development (RTD) activities oriented towards the automated recognition of some specific states of mind. Although the particular focus of our research is on the states of mind associated with negative emotions, mental strain and high cognitive effort, the CLAS dataset could offer an adequate support to research of a wider scope, such as general studies on attention assessment, cognitive load assessment, emotion recognition, as well as stress detection. The dataset consists of synchronized recordings of physiological signals, such as Electrocardiography (ECG), Plethysmography (PPG), ElectroDermal Activity (EDA), as well as accelerometer data, and metadata of 62 healthy volunteers, which were recorded while involved in three interactive tasks and two perceptive tasks. The interactive tasks aim to elicit different types of cognitive effort and included solving sequences of Math problems, Logic problems and the Stroop test. The perceptive tasks make use of images and audio-video stimuli, purposely selected to evoke emotions in the four quadrants of the arousal-valence space. The joint analysis of success rates in the interactive tasks and the information acquired through the questionnaire and the physiological recordings enables for a multifaceted evaluation of specific states of mind. These results are important for the advancement of research on efficient human-robot collaborations and general research on intelligent human-machine interaction interfaces.

Prolonged static sitting at the workplace is considered one of the main risks for the development of musculoskeletal disorders (MSDs) and adverse health effects. Factors such as poor posture and extended sitting are perceived to be a reason for conditions such as lumbar discomfort and lower back pain (LBP), even though the scientific explanation of this relationship is still unclear and raises disputes in the scientific community. The current study focused on evaluating the relationship between LBP and prolonged sitting in poor posture using photogrammetric images, postural angle calculation, machine learning models, and questionnaire-based self-reports regarding the occurrence of LBP and similar symptoms among the participants. Machine learning models trained with this data are employed to recognize poor body postures. Two scenarios have been elaborated for modeling purposes: scenario 1, based on natural body posture tagged as correct and incorrect, and scenario 2, based on incorrect body postures, corrected additionally by the rehabilitator. The achieved accuracies of respectively 75.3% and 85% for both scenarios reveal the potential for future research in enhancing awareness and actively managing posture-related issues that elevate the likelihood of developing lower back pain symptoms.

Prolonged computer-related work can be linked to musculoskeletal disorders (MSD) in the upper limbs and improper posture. In this regard, we report on developing resources supporting improper posture studies based on motion capture sensors. These resources were used to create a baseline detector for the automated detection of improper sitting postures, which was next used to evaluate the applicability of Hjorth’s parameters—Activity, Mobility and Complexity—on the specific classification task. Specifically, based on accelerometer data, we computed Hjorth’s time-domain parameters, which we stacked as feature vectors and fed to a binary classifier (kNN, decision tree, linear SVM and Gaussian SVM). The experimental evaluation in a setup involving two different keyboard types (standard and ergonomic) validated the practical worth of the proposed sitting posture detection method, and we reported an average classification accuracy of up to 98.4%. We deem that this research contributes toward creating an automated system for improper posture monitoring for people working on a computer for prolonged periods.

We present an adaptive feature selection method that makes use of Fisher’s discriminant ratio (FDR) with flexible threshold which is adjusted in a person-specific manner. The proposed method is shown to improve the detection of high-arousal negative-valence (HANV) conditions, based on two combinations of physiological signals (ECG+GSR and PPG+GSR). We validate the proposed method in an experimental setup aiming at the automated detection of HANV conditions evoked by audio-visual stimuli and picture stimuli. The experimental results support that the proposed method yields to an improvement of the classification accuracy of an SVM-based detector on average with 5.6%±0.6% in comparison with the traditional non-adaptive FDR-based feature selection using threshold 0.3, and with the full set of 39 features.

We present a three-step method for attribute selection that builds on person-independent and person-specific feature assessment stages. The first two steps aim to select a person-independent subset of attributes that are repeatedly selected for a large population of users. Next, this selection is intersect with a person-specific subset derived from the Fisher's separation criterion. As a result, we obtain a subset of attributes which is both task-specific and customized to the quality of data of each particular user. The proposed method was validated on the ASCERTAIN database in an experimental setup oriented towards high-arousal negative-valence detection based on physiological signals. The experimental results support that the proposed method offers advantage in terms of detection accuracy when compared to other subset selection strategies.