Modeling Engagement with Multimodal Multisensor Data: The Continuous Performance Test as an Objective Tool to Track Flow

Engagement is one of the most important factors in determining successful outcomes and deep learning in students. Existing approaches to detect student engagement involve periodic human observations that are subject to inter-rater reliability. Our solution uses real-time multimodal multisensor data labeled by objective performance outcomes to infer the engagement of students. The study involves four students with a combined diagnosis of cerebral palsy and a learning disability who took part in a 3-month trial over 59 sessions. Multimodal multisensor data were collected while they participated in a continuous performance test. Eye gaze, electroencephalogram, body pose, and interaction data were used to create a model of student engagement through objective labeling from the continuous performance test outcomes. In order to achieve this, a type of continuous performance test is introduced, the Seek-X type. Nine features were extracted including high-level handpicked compound features. Using leave-one-out cross-validation, a series of different machine learning approaches were evaluated. Overall, the random forest classification approach achieved the best classification results. Using random forest, 93.3% classification for engagement and 42.9% accuracy for disengagement were achieved. We compared these results to outcomes from different models: AdaBoost, decision tree, k-Nearest Neighbor, naïve Bayes, neural network, and support vector machine. We showed that using a multisensor approach achieved higher accuracy than using features from any reduced set of sensors. We found that using high-level handpicked features can improve the classification accuracy in every sensor mode. Our approach is robust to both sensor fallout and occlusions. The single most important sensor feature to the classification of engagement and distraction was shown to be eye gaze. It has been shown that we can accurately predict the level of engagement of students with learning disabilities in a real-time approach that is not subject to inter-rater reliability, human observation or reliant on a single mode of sensor input. This will help teachers design interventions for a heterogeneous group of students, where teachers cannot possibly attend to each of their individual needs. Our approach can be used to identify those with the greatest learning challenges so that all students are supported to reach their full potential.

Enhancement of Stereo Video Pairs Using SDNs To Aid In 3D Reconstruction

This paper presents the results of enhancing images from a left and right stereo pair in order to increase the resolution of a 3D representation of a scene generated from that same pair. A new neural network structure known as a Self Delaying Dynamic Network (SDN) has been used to perform the enhancement. The advantage of SDNs over existing techniques such as bicubic interpolation is their ability to cope with motion and noise effects. SDNs are used to generate two high resolution images, one based on frames taken from the left view of the subject, and one based on the frames from the right. This new high resolution stereo pair is then processed by a disparity map generator. The disparity map generated is compared to two other disparity maps generated from the same scene. The first is a map generated from an original high resolution stereo pair and the second is a map generated using a stereo pair which has been enhanced using bicubic interpolation. The maps generated using the SDN enhanced pairs match more closely the target maps. The addition of extra noise into the input images is less problematic for the SDN system which is still able to out perform bicubic interpolation.

Enhancing Multi-Frame Images Using Self-Delaying Dynamic Networks

This paper presents the use of a newly created network structure known as a Self-Delaying Dynamic Network (SDN) to create a high resolution image from a set of time stepped input frames. These SDNs are non-recurrent temporal neural networks which can process time sampled data. SDNs can store input data for a lifecycle and feature dynamic logic based connections between layers. Several low resolution images and one high resolution image of a scene were presented to the SDN during training by a Genetic Algorithm. The SDN was trained to process the input frames in order to recreate the high resolution image. The trained SDN was then used to enhance a number of unseen noisy image sets. The quality of high resolution images produced by the SDN is compared to that of high resolution images generated using Bi-Cubic interpolation. The SDN produced images are superior in several ways to the images produced using Bi-Cubic interpolation.