Search

Human activity recognition (HAR) based on wearable sensors has attracted significant research attention in recent years due to its advantages in availability, accuracy, and privacy-friendliness. HAR baseline model is essentially a general-purpose classifier trained to recognized multiple activity patterns of most user types. It provides the input for subsequent steps of model personalization. Training a good baseline model is of fundamental importance because it has significant impacts on the ultimate HAR accuracy. In practice, baseline model training in HAR is a non-trivial problem that faces two challenges: insufficient training data and biased training data. This paper proposes a novel baseline model training scheme to tackle the two challenges using Deep InfoMax (DIM)-based unsupervised feature extraction and Broad Learning System (BLS)-based incremental learning, respectively. Experimental results demonstrate that the proposed scheme outperform conventional methods in terms of overall accuracy, computational efficiency, and the ability to adapt to dynamic scenarios with changing data characteristics.

The accurate depiction and understanding of the influence mechanism of residents’ dependence on public transport (RDPT) is an important foundation for increasing the proportion of green trips and alleviating urban traffic congestion. To explore the influence mechanism of RDPT, this paper extends the theory of planned behavior (TPB) by introducing objective factors of attributes, environment, and travel characteristics. The agglomerative nesting clustering algorithm and multiple structural equation modeling (SEM) are then developed to identify the RDPT levels and analyze the influence relationships between the integrated influencing factors and RDPT based on travel survey data. The results indicate that the objective variables have indirect impacts on RDPT by influencing psychological variable attitudes, subjective norms and perceived control and travel intention. The residents’ self-selection (RSS) effects of different clusters are all detected under normal conditions, and the environment on RDPT is still significant after controlling for this effect. The findings reflect that the influence mechanisms of the three SEMs for different clusters are distinct, unlike those of the baseline model, and distinctive observation variables have dissimilar explanatory abilities for the travel intention of different residents. Therefore, some beneficial policy implications are proposed for residents, especially those who retain low and relatively low public transport dependence levels, to increase public transport usage while reduce the car usage based on the significant findings.

Automated ultrasonic signal classification systems are often utilized for the recognition of a large number of ultrasonic signals in engineering materials. Existing defect classification methods are mainly image-based and serve to extract features for various defects. In this paper, we propose a novel detection baseline model based on a fully convolution network (FCN) and gated recurrent unit (GRU) to classify ultrasonic signals from flawed 3D braided composite specimens with debonding defects. In the proposed algorithm, the proposed Gated Recurrent Unit Fully Convolutional Network (GRU-FCN) is used to extract temporal characteristics of ultrasonic signals; the GRU module is key to enhancing the performance of FCNs. Experimental results on an in-house dataset indicated that the proposed model performs very well against all baselines. We also developed a scheme to interpret the relationship between A-scan and C-scan images and a 3D depth model representation to visualize the location information of defects.