Mixed graph convolution and residual transformation network for skeleton-based action recognition

Action recognition based on a human skeleton is an extremely challenging research problem. The temporal information contained in the human skeleton is more difficult to extract than the spatial information. Many researchers focus on graph convolution networks and apply them to action recognition. In this study, an action recognition method based on a two-stream network called RNXt-GCN is proposed on the basis of the Spatial-Temporal Graph Convolutional Network (ST-GCN). The human skeleton is converted first into a spatial-temporal graph and a SkeleMotion image which are input into ST-GCN and ResNeXt, respectively, for performing the spatial-temporal convolution. The convolved features are then fused. The proposed method models the temporal information in action from the amplitude and direction of the action and addresses the shortcomings of isolated temporal information in the ST-GCN. The experiments are comprehensively performed on the four datasets: 1) UTD-MHAD, 2) Northwestern-UCLA, 3) NTU RGB-D 60, and 4) NTU RGB-D 120. The proposed model shows very competitive results compared with other models in our experiments. On the experiments of NTU RGB?+?D 120 dataset, our proposed model outperforms those of the state-of-the-art two-stream models.

     

A joint model based on interactive gate mechanism for spoken language understanding

Applied Intelligence - Slot filling and intent detection are two important tasks in a spoken language understanding (SLU) system, it is becoming a tendency that two tasks are jointing learn in SLU....     

Synchronously tracking entities and relations in a syntax-aware parallel architecture for aspect-opinion pair extraction

Aspect-Opinion Pair Extraction (AOPE) task aims to capture each aspect with its corresponding opinions in user reviews. Entity recognition and relation detection are two fundamental subtasks of AOPE. Although recent works take interaction into account, the two subtasks are still relatively independent during calculation. Furthermore, since AOPE task has not been formally proposed for a long time, syntactic information does not attract much attention in the current deep learning models for AOPE. In this paper, we propose a model for Synchronously Tracking Entities and Relations (STER) to deal with AOPE. Specifically, we design a network consisting of a bank of gated RNNs, where we can track all entities of a review sentence in parallel. STER utilizes three features, i.e., context, syntax and relation, to learn the representation of each tracked entity and calculate the correlated degree between all entities synchronously at each time step. The entity representation and the correlated degree are highly dependent during calculation. Finally, they will be used for entity recognition and relation detection, respectively. Therefore, in STER, the two subtasks of AOPE can achieve sufficient interaction, which enhances their mutual heuristic effect heavily. To verify the effectiveness and adaptiveness of our model, we conduct experiments on two annotation versions of SemEval datasets. The results demonstrate that STER not only achieves advanced performances but adapts to different annotation strategies well.

     

Enhanced semantic representation learning for implicit discourse relation classification

Applied Intelligence - Implicit discourse relation classification is one of the most challenging tasks in discourse parsing. Without connectives as linguistic clues, classifying discourse relations...     

Analysis of Internet Financial Risks Based on Deep Learning and BP Neural Network

Computational Economics - The study aims to analyze and forecast Internet financial risks based on the model based on deep learning and the Back Propagation Neural Network (BPNN). First, the theory...     

A novel approach of data race detection based on CNN-BiLSTM hybrid neural network

Neural Computing and Applications - Existing data race detection approaches based on deep learning are suffering from the problems of unique feature extraction and low accuracy. To this end, this...     

Simulation analysis and experimental verification of coal suction characteristics of the new railway tunnel fallen coal dust collection device北大核心CSCD

Due to air turbulence, large areas of coal will fall when the special coal-transportation trains pass the tunnel exits and entrances. Aiming at the problems of low efficiency and high cost of manual cleaning for long distance coal cleaning in the tunnel, a new railway tunnel fallen coal dust collection device which was composed of a main conveying coal feeding pipe and multiple branch pipes of coal suction was designed. It was used to clean the small particles and lightweight railway tunnel fallen coal. Firstly, the gas-solid two-phase flow model based on the Euler-Lagrange approach for the design of the main conveying coal feeding pipe was established in the coal conveying pipelines. Secondly, the effect of the coal particles' incident angle and multiple branch pipe spacing on the main coal conveying pipe flow field, which was based on Fluent finite element simulation software, was studied. What was more, the optimal angle of incidence and the optimal value of the number of branch coal suction pipe, which was installed on the main conveying pipe, were analyzed. Finally, the finite element simulation was verified by field test. Simulation and experimental results showed that it was more conducive to the railway tunnel fallen coal transportation when coal particles' incident angle was less than 45° and the branch pipe spacing was in the vicinity of 750 mm. For that when incident angle was less than 45°, the main conveying coal pipe pressure-drop became weaker and particle flow could obtain large horizontal transport velocity. And when the branch pipe spacing was in the vicinity of 750 mm, the horizontal transport velocity had a smaller fluctuation range and the transportation of coal was larger than that of the other groups. The research results are of great significance to improve the structure of the main conveying coal pipe, increase the efficiency of tunnel coal conveying and optimize the railway tunnel coal dust collection device.     

Effect of Sintering Temperature on Magnetic and Electrical Properties of GdBaCo<Subscript>2</Subscript><Emphasis Type="Italic">O</Emphasis><Subscript>5 + <Emphasis Type="Italic">δ</Emphasis></Subscript>

In this study, we report the results of an investigation into the sintering temperature dependence of magnetic and transport properties for GdBaCo₂ O _{5 + δ} synthesized through a sol-gel method. The lowering of sintering temperature leads to the increase of oxygen content and the reduction of grain size. The increase of oxygen content results in the enhancement of magnetic interactions and the weakening of Coulomb repulsion effect, while the reduction of grain size improves the magnetoresistance effect. Metal-insulator transition accompanied with spin-state transition is observed in all samples.