A deep neural network model for multi-view human activity recognition.

Multiple cameras are used to resolve occlusion problem that often occur in single-view human activity recognition. Based on the success of learning representation with deep neural networks (DNNs), recent works have proposed DNNs models to estimate human activity from multi-view inputs. However, currently available datasets are inadequate in training DNNs model to obtain high accuracy rate. Against such an issue, this study presents a DNNs model, trained by employing transfer learning and shared-weight techniques, to classify human activity from multiple cameras. The model comprised pre-trained convolutional neural networks (CNNs), attention layers, long short-term memory networks with residual learning (LSTMRes), and Softmax layers. The experimental results suggested that the proposed model could achieve a promising performance on challenging MVHAR datasets: IXMAS (97.27%) and i3DPost (96.87%). A competitive recognition rate was also observed in online classification.

Identifying autism spectrum disorder symptoms using response and gaze behavior during the Go/NoGo game CatChicken.

Previous studies have found that Autism Spectrum Disorder (ASD) children scored lower during a Go/No-Go task and faced difficulty focusing their gaze on the speaker's face during a conversation. To date, however, there has not been an adequate study examining children's response and gaze during the Go/No-Go task to distinguish ASD from typical children. We investigated typical and ASD children's gaze modulation when they played a version of the Go/No-Go game. The proposed system represents the Go and the No-Go stimuli as chicken and cat characters, respectively. It tracks children's gaze using an eye tracker mounted on the monitor. Statistically significant between-group differences in spatial and auto-regressive temporal gaze-related features for 21 ASD and 31 typical children suggest that ASD children had more unstable gaze modulation during the test. Using the features that differ significantly as inputs, the AdaBoost meta-learning algorithm attained an accuracy rate of 88.6% in differentiating the ASD subjects from the typical ones.