The Sensor-based Human Activity Recognition (SHAR) task seeks to recognize signals collected by various sensors embedded in intelligent devices to assist people in their daily lives. Motivated by the success of deep learning, many researchers are studying combining deep learning with SHAR. The key to implementing sensor-based human activity recognition with deep learning lies in facilitating model performance and maintaining efficiency when the model is performed on resource-constrained devices. To address this challenge, we present an effective sensor signal representation method, termed the Temporal-Spatial Dynamic Convolutional (TS-DyConv) network, to recognize human activity. TS-DyConv aims to dynamically learn the convolutional kernels weighted with attention generated along the temporal and spatial kernel spaces. This way, the TS-DyConv can diversify the kernel to enhance the sensor signal’s recognition capabilities without raising complexity and maintaining efficiency. Extensive experiments conducted on three benchmark SHAR datasets, e.g., OPPORTUNITY, PAMAP2, and USC-HAD, demonstrate the superiority of our method over the deep learning baselines and existing SHAR works.

Recognizing and understanding various objects in visual information is paramount for ensuring safe and efficient autonomous navigation, especially in challenging environmental conditions. However, relying solely on single-modal data to perceive information, such as visible images, can compromise the quality and reliability of the extracted information, posing potential risks to autonomous driving systems. To address this challenge, we present a novel fusion method based on mutual information theory in semantic segmentation tasks for secure and efficient autonomous vehicles. The proposed method involves two primary components, i.e., multispectral fusion representation module (FRM) and semantic segmentation module (SSM). To optimize the FRM, we establish a unified quality representation for feature fusion by incorporating an image restoration mechanism, enhancing autonomous driving systems' overall performance and adaptability in complex environments. Meanwhile, we employ mutual information maximization to capture the interimage relations among pixels from the same semantic content, which is achieved by leveraging the semantic representation learned by the SSM in a high-dimensional feature space. Experimental results and comparisons with famous fusion approaches and segmentation models on four public datasets validate our method's effectiveness, robustness, and overall superiority.

The Sensor-based Human Activity Recognition (SHAR) task aims to detect and analyze signals captured by various sensors embedded in intelligent devices to assist people’s daily lives, which inherently have variations, spatial inconsistency, and noise characteristics. Inspired by the success of deep learning, many researchers have designed various neural network architectures to improve the SHAR model’s performance. Albeit effective, these manners are not optimal due to ignoring the imbalance and difficulty of inter- and intra-class of sensor inherently characteristic, which proposes the challenge for resource-constrained devices. To address this challenge, we present a simple and effective optimization strategy for sensor signal processing, including Dynamic Weight Optimization (DWO) and Smoothness Constraint (SC) strategies, termed DWOSC, which provide a framework to optimize loss function and enhance activity recognition performance. The DWO aims to address the class imbalance and hard-to-recognize samples. The SC seeks to maintain the smoothness of decision boundaries for complex samples. This way, the proposed approach can achieve a more stable convergence of the model and further improve the SHAR task’s performance without raising complexity. Extensive experiments conducted on three benchmark SHAR datasets, e.g., OPPORTUNITY, USC-HAD, and UniMib, demonstrate the superiority of our method over the deep learning baselines and existing SHAR works.