Automatic emotion recognition from the analysis of body movement has tremendous potential to revolutionize virtual reality, robotics, behavior modeling, and biometric identity recognition domains. A computer system capable of recognizing human emotion from the body can also significantly change the way we interact with the computers. One of the significant challenges is to identify emotion-specific features from a vast number of descriptors of human body movements. In this paper, we introduce a novel two-layer feature selection framework for emotion classification from a comprehensive list of body movement features. We used the feature selection framework to accurately recognize five basic emotions: happiness, sadness, fear, anger, and neutral. In the first layer, a unique combination of Analysis of Variance (ANOVA) and Multivariate Analysis of Variance (MANOVA) was utilized to eliminate irrelevant features. In the second layer, a binary chromosome-based genetic algorithm was proposed to select a feature subset from the relevant list of features that maximizes the emotion recognition rate. Score and rank-level fusion were applied to further improve the accuracy of the system. The proposed system was validated on proprietary and public datasets, containing 30 subjects. Different action scenarios, such as walking and sitting actions, as well as an action-independent case, were considered. Based on the experimental results, the proposed emotion recognition system achieved a very high emotion recognition rate outperforming all of the state-of-the-art methods. The proposed system achieved recognition accuracy of 90.0% during walking, 96.0% during sitting, and 86.66% in an action-independent scenario, demonstrating high accuracy and robustness of the developed method.

Motion capture sensor-based gait emotion recognition is an emerging sub-domain of human emotion recognition. Its applications span a variety of fields including smart home design, border security, robotics, virtual reality, and gaming. In recent years, several deep learning-based approaches have been successful in solving the Gait Emotion Recognition (GER) problem. However, a vast majority of such methods rely on Deep Neural Networks (DNNs) with a significant number of model parameters, which lead to model overfitting as well as increased inference time. This paper contributes to the domain of knowledge by proposing a new lightweight bi-modular architecture with handcrafted features that is trained using a RMSprop optimizer and stratified data shuffling. The method is highly effective in correctly inferring human emotions from gait, achieving a micro-mean average precision of 0.97 on the Edinburgh Locomotive Mocap Dataset. It outperforms all recent deep-learning methods, while having the lowest inference time of 16.3 milliseconds per gait sample. This research study is beneficial to applications spanning various fields, such as emotionally aware assistive robotics, adaptive therapy and rehabilitation, and surveillance.

Over the past decade, gait recognition had gained a lot of attention in various research and industrial domains. These include remote surveillance, border control, medical rehabilitation, emotion detection from posture, fall detection, and sports training. The main advantages of identifying a person by their gait include unobtrusiveness, acceptance, and low costs. This paper proposes a convolutional neural network KinectGaitNet for Kinect-based gait recognition. The 3D coordinates of each of the body joints over the gait cycle are transformed to create a unique input representation. The proposed KinectGaitNet is trained directly using the 3D input representation without the necessity of the handcrafted features. The KinectGaitNet design allows avoiding gait cycle resampling, and the residual learning method ensures high accuracy without the degradation problem. The proposed deep learning architecture surpasses the recognition performance of all state-of-the-art methods for Kinect-based gait recognition by achieving 96.91% accuracy on UPCV and 99.33% accuracy on the KGB dataset. The method is the first, to the best of our knowledge, deep learning-based architecture that is based on a unique 3D input representation of joint coordinates. It achieves performance higher than previous traditional and deep learning methods, with fewer parameters and shorter inference time.