This paper proposes a human activity recognition (HAR) method for frequency-modulated continuous wave (FMCW) radar sensors. The method utilizes a multi-domain feature attention fusion network (MFAFN) model that addresses the limitation of relying on a single range or velocity feature to describe human activity. Specifically, the network fuses time-Doppler (TD) and time-range (TR) maps of human activities, resulting in a more comprehensive representation of the activities being performed. In the feature fusion phase, the multi-feature attention fusion module (MAFM) combines features of different depth levels by introducing a channel attention mechanism. Additionally, a multi-classification focus loss (MFL) function is applied to classify confusable samples. The experimental results demonstrate that the proposed method achieves 97.58% recognition accuracy on the dataset provided by the University of Glasgow, UK. Compared to existing HAR methods for the same dataset, the proposed method showed an improvement of about 0.9–5.5%, especially in the classification of confusable activities, showing an improvement of up to 18.33%.

Non-contact vital sign monitoring, based on mm-wave Frequency Modulated Continuous Wave (FMCW) radar, can be widely applied in elderly care, clinical diagnoses, and in-vehicle occupant detection. The radar can monitor the Doppler signal reflected by the chest displacement, then analyze the phase change of echo to get the respiration and heartbeat signals. This paper proposes a frequency-tracking method based on an improved Birth-Death strategy to calculate vital sign rate in real-time. There are two main steps to estimate vital sign rate from frames within a second. Firstly, the M-Rife algorithm is employed to interpolate the frequency of each frame. Secondly, find the matching frequency points between the adjacent frames in a second by using an improved Birth-Death strategy and calculate the frequency ratio for each matching points link. Then, choose the longest link whose initial frequency is closer to the frequency at the last second. The value of the link tail will be a vital sign rate at the current second. The proposed method is simulated in a public non-contact vital dataset to verify its feasibility and accuracy. To further demonstrate the method, this paper carried out a series of experiments. The experiment results show that the method has an accuracy rate of 98% for the heartbeat rate in real-time, and reduces the effect of clutter and respiratory harmonics on the heartbeat rate.

Weakly supervised video anomaly detection aims to detect anomalous events with only video-level labels. In the absence of boundary information for anomaly segments, most existing methods rely on multiple instance learning. In these approaches, the predictions for unlabeled video snippets are guided by the classification of labeled untrimmed videos. However, these methods do not account for issues such as video blur and visual occlusion, which can hinder accurate anomaly detection. To address these issues, we propose a novel weakly supervised video anomaly detection method that fuses multimodal and multiscale features. Firstly, RGB and optical flow snippets are input into pre-trained I3D to extract appearance and motion features. Then, we introduce an Attention De-redundancy (AD) module, which employs an attention mechanism to filter out task-irrelevant redundancy in these appearance and motion features. Next, to mitigate the effects of video blurring and visual occlusion, we propose a Multi-scale Feature Learning module. This module captures long-term and short-term temporal dependencies among video snippets to provide global and local guidance for blurred or occluded video snippets. Finally, to effectively utilize the discriminative features of different modalities, we propose an Adaptive Feature Fusion module. This module adaptively fuses appearance and motion features based on their respective feature weights. Extensive experimental results demonstrate that our proposed method outperforms mainstream unsupervised and weakly supervised methods in terms of AUC. Specifically, our proposed method achieves 97.00% AUC and 85.31% AUC on two benchmark datasets, i.e., ShanghaiTech and UCF-Crime, respectively.

This paper investigates the prospective n-point (PnP) problem and provides an effective, dependable, and fast optimal solution. The challenging pose estimation problem is changed into the ideal circumstance for solving the transition matrix by estimating the position and orientation of the camera based on known reference points. This method meets both the geometric optimality and the statistical optimal solution by accounting for observation and propagation uncertainty in the solution process. In addition, in order to further optimize the error caused by the mapping process, this paper introduces the L-BFGS algorithm for optimal processing. It saves time and cost because it can iterate the optimal transition matrix and systematic error synchronously. In this paper, the algorithm simulation tests are carried out in the standard case, the quasi-singular case, and the planar case. The results of the experiments demonstrate that the strategy suggested in this paper can successfully address the issue of camera position estimation. Compared with the existing advanced technology, the accuracy is improved by about 8% - 13%, and the time cost is low.