Abstract:Human activity recognition (HAR) technology based on wearables has received increasing attention in recent years. The traditional methods have used hand-crafted features to recognize human activities, resulting in shallow feature extraction. With the development of deep learning, an increasing number of researchers have focused on studying deep learning methods. To achieve higher recognition accuracy, the majority of the current HAR research involves multisource and multimodal sensors (MMSs) data. However, due to the limitations in the receptive fields of single-dimensional convolutional kernels, these networks are still infeasible for extracting spatiotemporal features. In this study, a multidimensional parallel convolutional connected (MPCC) deep learning network based on MMS data for HAR is proposed that fully utilizes the advantages of multidimensional convolutional kernels. Moreover, multiscale residual convolutional squeeze-and-excitation (MRCSE) modules are proposed to enrich the diversity of feature information by combining squeeze-and-excitation (SE) blocks. A daily home activity (DHA) data set is constructed based on the requirements for HAR in certain scenarios, such as smart home, and we conduct experiments on the optimal combination of sensor locations on the DHA data set according to a weighted $\text{F}1~({\mathrm{ F}}_{\mathrm{ W}})$ -score. Both tenfold and leave-one-subject-out (LOSO) cross-validations (CVs) are used to evaluate the performance of the proposed network. The MPCC-MRCSE network achieves ${\mathrm{ F}}_{\mathrm{ W}}$ -scores of 98.33% and 95.42% on the physical activity monitoring for aging people (PAMAP2) and OPPORTUNITY data sets using tenfold CVs, respectively, and achieves ${\mathrm{ F}}_{\mathrm{ W}}$ -scores of 81.47% on the PAMAP2 when applying an LOSO CV.

Human activity recognition method combined with region HOF and dictionary learning

Design method for action recognition applied to senile apartment

Human Action Recognition Using Deep Learning Methods.

Discovery and Recognition of Emerging Human Activities Using a Hierarchical Mixture of Directional Statistical Models.

Context-Associative Hierarchical Memory Model for Human Activity Recognition and Prediction

Human activity recognition for video surveillance

Human Activity Recognition based on Dynamic Spatio-Temporal Relations

Activity Recognition Based On Hidden Markov Models

Combining Multi-scale Directed Depth Motion Maps and Log-Gabor Filters for Human Action Recognition

Human Action Recognition by Using Polyhedron Model-Based Spatio-Temporal Gradient Descriptor

Human Activities Prediction by Learning Combinatorial Sparse Representations.

Human Action Recognition Based on HOIRM Feature Fusion and AP Clustering BOW.

A Hierarchical Spatio-Temporal Model for Human Activity Recognition.

Human Action Recognition Algorithm Based on Spatial Temporal Depth Feature

Activity Recognition Via Feature Decomposition

A Multi-dimensional Parallel Convolutional Connected Network Based on Multi-source and Multi-modal Sensor Data for Human Activity Recognition

Oriented Gradients for Human Action Recognition.

Human action recognition based on Non-linear SVM Decision Tree

A Multidimensional Parallel Convolutional Connected Network Based on Multisource and Multimodal Sensor Data for Human Activity Recognition

Recognizing Human Activities Using Non-linear SVM Decision Tree

Recognize Human Activities From Multi-Part Missing Videos