Our goal is to enhance the effectiveness of teaching and learning by computing students affective, since affective indicates learning states of students. We have built a facial expression database which includes more than 300, 000 labeled facial images( these faces are labeled into 15 AUs according to FACS (Facial Action Coding System). We have been exploring how to analyze emotions by deep learning approaches on both public databases and on our own educational emotion database.
Gestures, for example, hand-raising, convey important information of students learning states in real classrooms. Gesture recognition aims at improving the teaching quality by giving hints about students’ learning states. We have constructed a gesture recognition database of more than 110,000 samples. We have been elaborately designing deep neural network to improve performance of gesture recognition in real classrooms.
Tensor, as a generalization of matrix, is a multidimensional array. Tensor representation of multidimensional data can preserve the intrinsic spatial structures, which has been demonstrated to be useful for numerous applications, such as hyperspectral image denoising, color image recovery, and image clustering. Due to the redundancy of high dimensional data, we focus on the tensor sparse representations, including new tensor sparse models and algorithms. Especially, we are interested in the tensor representation under the circular algebra system. Some definitions have similar formulations and properties as those in classical linear algebra system, such as tensor product (t-product), tensor linear combination (t-linear combination), tensor singular value decomposition (t-SVD).
The Mobile Learning system is a cutting-edge mobile learning system that can deliver live broadcasts of real-time classroom teaching to students with mobile devices. The system allows students to customize means of content-reception, based on when and where they tune into the broadcast. The system also supports short text-messaging and instant polls. Through these venues, students can ask questions and make suggestions in real time. The instructor can address them immediately. The evaluation from a formal implementation of the system in a blended English classroom of 1,000 students (with about 800 being online) reveals that Mobile Learning activities can much better engage students in the learning process. Students in this class changed from passive learners to truly engaged learners who are behaviorally, intellectually, and emotionally involved in their learning tasks.