Good sleep is very important for everyone to protect physical and mental health. People's sleep behavior at night reflects their sleep status. In this work, we propose a method to detect people's sleep behavior at night by adopting Pseudo-3D (P3D) convolution neural network with attention mechanism.In particular, we propose a new structure, which integrates Squeeze-and-Excitation (SE) blocks into P3D blocks, named P3D-Attention. For the input video, we use P3D blocks to extract spatial-temporal features, and use SE blocks to pay more attentions to the important channel features. The proposed network is tested on the Sleep Action (SA) dataset, which consists of five different actions, namely turn over, get up, fall off bed, play mobile phone, and normal sleep. Experimental results show that the proposed network achieves reasonably good detection results, and the accuracy rate on the test set can reach 90.67%. Compared with 3D convolutional neural networks (C3D), our proposed network can increase the accuracy by about 6% with only 1/6 model parameter size, and achieves an average prediction speed about 1.75 item/s. Compared with the residual spatiotemporal convolution network (R(2+1)D), our proposed network can increase the accuracy rate by about 1.5% with less than 1/2 model parameter size.

No abstract

Aspect information mining from user comments has become an important means to improve the performance of recommendation systems (RSs). This is because aspect information in comments is fine-grained and tends to reflect the interactions and preferences of users over items in multiple dimensions. These interactions are different from ratings, which are often explicit and linear. Most current RSs based on aspect information learn the contribution of explicit interactions of aspects in a linear manner, while ignoring the implicit features and non-linear interactions of aspects. Since Chinese grammar is greatly different with English grammar, there are few recommendation models based on Chinese movie comment aspects. In this work, we propose an architecture, named aspect-based neural collaborative filtering (ANCF), to extract comment aspect terms based on rules formulated in Chinese dependency parsing. The proposed ANCF integrates a generalized tensor factorization and a tensorized multi-layer perceptrons into the neural network to capture user-item-aspect interactions in a mixed linear and nonlinear way. The aspect potential interaction vector and the actual interaction vector are layered and fused into tensor processing, which can reduce the tensor sparsity and solve the cold start problem of collaborative filtering to a certain extent. Performance results show that the proposed model outperforms some of the traditional ones in terms of recommendation accuracy and effectiveness.INDEX TERMS Aspect-aware implicit interactions, collaborative filtering, tensor factorization, deep learning, Chinese comment mining

No abstract