Smart homes have become central in the sustainability of buildings. Recognizing human activity in smart homes is the key tool to achieve home automation. Recently, two-stream Convolutional Neural Networks (CNNs) have shown promising performance for video-based human action recognition. However, such models cannot act directly on the 3D skeletal sequences due to its limitation to the 2D image video inputs. Considering the powerful effect of 3D skeletal data for describing human activity, in this study, we present a novel method to recognize the skeletal human activity in sustainable smart homes using a CNN fusion model. Our proposed method can represent the spatiotemporal information of each 3D skeletal sequence into three images and three image sequences through gray value encoding, referred to as skeletal trajectory shape images (STSIs) and skeletal pose image (SPI) sequences, and build a CNNs’ fusion model with three STSIs and three SPI sequences as input for skeletal activity recognition. Such three STSIs and three SPI sequences are, respectively, generated in three orthogonal planes as complementary to each other. The proposed CNN fusion model allows the hierarchical learning of spatiotemporal features, offering better action recognition performance. Experimental results on three public datasets show that our method outperforms the state-of-the-art methods.

This paper addresses the problem of detecting money laundering in the Bitcoin network. Money laundering is the process of handling the proceeds of crime to conceal their illegal source, these illicit transactions have complex features, similar to those of legal transactions. It is well known that transactions can be represented as topological graph structure data, and many GCN-based methods have been developed for Anti-Money Laundering (AML) tasks. However, existing methods have not performed as well in dynamically assigning weights to neighboring nodes and extracting information from global nodes in the Bitcoin network. Therefore, we identify three major challenges: Firstly, GCNs can be misled by concealed illegal transactions due to uniform node representation weights. Secondly, current node-level GCNs cannot handle varied methods of concealing illegal transactions because they fail to extract global information. Thirdly, the costliness of data labelling necessitates the effective use of limited but rich domain-specific labelled data. To address these challenges, we propose the Transformer-enhanced Graph Attention Network (TFGAT) with a Global-Local Attention Mechanism (GLATM) that uses Transformers to extract global information and selectively focus on local information from connected nodes. Due to the limited availability of labelled data from expensive data labelling processes, we introduce a Deep Cyclic Pseudo-Label Updating Mechanism (DCPLU) to enhance data distribution and model robustness, which does not rely on manifold structure or Euclidean distance assumptions. DCPLU can enhance model performance while preserving the model’s existing parameters, enabling it to maintain its current faster response time in the application scenario. Experimental results show that our methods outperform existing models across various metrics.

This paper addresses the problem of predicting human actions in depth videos. Due to the complex spatiotemporal structure of human actions, it is difficult to infer ongoing human actions before they are fully executed. To handle this challenging issue, we first propose two new depth-based features called pairwise relative joint orientations (PRJOs) and depth patch motion maps (DPMMs) to represent the relative movements between each pair of joints and human-object interactions, respectively. The two proposed depth-based features are suitable for recognizing and predicting human actions in real-time fashion. Then, we propose a regression-based learning approach with a group sparsity inducing regularizer to learn action predictor based on the combination of PRJOs and DPMMs for a sparse set of joints. Experimental results on benchmark datasets have demonstrated that our proposed approach significantly outperforms existing methods for real-time human action recognition and prediction from depth data.

Very recently, the BERT graph convolutional network (BertGCN) model has attracted much attention from researchers due to its good text classification performance. However, just using original documents in the corpus to construct the topology of graphs for GCN-based models may lose some effective information. In this paper, we focus on sentiment classification, an important branch of text classification, and propose the multistream BERT graph convolutional network (MS-BertGCN) for sentiment classification based on cross-document learning. In the proposed method, we first combine the documents in the training set based on within-class similarity. Then, each heterogeneous graph is constructed using a group of combinations of documents for the single-stream BertGCN model. Finally, we construct multistream-BertGCN (MS-BertGCN) based on multiple heterogeneous graphs constructed from different groups of combined documents. The experimental results show that our MS-BertGCN model outperforms state-of-the-art methods on sentiment classification tasks.