Search

Facial emotion recognition (FER) has been extensively researched over the past two decades due to its direct impact in the computer vision and affective robotics fields. However, the available datasets to train these models include often miss-labelled data due to the labellers bias that drives the model to learn incorrect features. In this paper, a facial emotion recognition system is proposed, addressing automatic face detection and facial expression recognition separately, the latter is performed by a set of only four deep convolutional neural network respect to an ensembling approach, while a label smoothing technique is applied to deal with the miss-labelled training data. The proposed system takes only 13.48 ms using a dedicated graphics processing unit (GPU) and 141.97 ms using a CPU to recognize facial emotions and reaches the current state-of-the-art performances regarding the challenging databases, FER2013, SFEW 2.0, and ExpW, giving recognition accuracies of 72.72%, 51.97%, and 71.82% respectively.

Security is a major concern in Electronic Payment (E-Payment) systems. Usually, these systems are protected against illegal users, so-called hackers, by different means, such as Personal identification numbers (PINs), passwords, cards, etc. However, these hackers may manage to bypass this protection by having recourse to different strategies. Many techniques have been proposed to counter hacking attempts; however, there are still situations where an illegal user may succeed to access the E-payment system easily by stealing from a legal user its payment card. The use of Artificial Intelligence methods for face authentication, like deep learning, has made facial biometry a highly developing and accurate technology, especially in the past decade. In this paper, we propose the joint use of deep learning-based facial biometry and RFID cards to reinforce the security of an E-Payment system. By doing so, we ensure that a user should be physically present carrying his RFID card to be able to access the E-Payment system. We have tested three deep learning-based face authentication models and validated them on MUCT and CASIA Face-V5 datasets, to choose the most suitable one for our proposed secured E-Payment system, obtaining top verification rates of 99.90% and 99.26%, respectively. Two versions of this system are proposed; in the first version, which is based on a Personnel Computer (PC) and a Raspberry card, face authentication is implemented in a PC and the control of the RFID reader is performed by a Raspberry Pi 3, whereas in the second version, which may be considered as an embedded system, all the job is accomplished by the Raspberry Pi.

Biometric technologies have been widely used in recent years for authentication and security purposes. Face recognition is one of the important techniques for its simple way to obtain the subject samples without being intrusive. However, this technology has been employed mostly using the visible spectrum only, which suffers from some limitations such as illumination change, facial expression and pose variations. The infrared spectrum offers some advantages over the visible spectrum, mainly the robustness to light change. In this paper, we propose a new multispectral framework that uses both infrared and visible spectra with an optimization based on a new proposed feature selection algorithm and the PSO-SVM. Experimental tests were conducted on IRIS OTCBVS Thermal/Visible and CSIST Lab 2 Databases. The obtained results clearly demonstrate the effectiveness of our new framework compared to a mono-spectral face recognition system.

Recognizing emotions in controlled conditions, based on facial expressions, has achieved high accuracies in the past years. This is still a challenging task for robots working in real-world scenarios due to different factors such as illumination, pose variation or occlusions. One of the next barriers of science is to give sociable robots the ability to fully engage in emotional interactions with users. In this paper a real-time emotion recognition system using a YOLO-based facial detection system and an ensemble CNN for sociable robots, is proposed. Experiments have been carried out on the most challenging database, FER 2013, giving a performance of 72.47% on test sets, achieving current standards.