Human Body Contour Research Articles

Three-Dimensional View Relationship-Based Context-Aware Emotion Recognition.

Context-aware emotion recognition (CAER) leverages comprehensive scene information, including facial expressions, body postures, and contextual background. However, current studies predominantly rely on facial expressions, body postures, and global contextual features; the interaction between the agents (target individuals) and other objects in the scene is usually absent or incomplete. In this article, a three-dimensional view relationship-based CAER (TDRCer) method is proposed, which comprises two branches: the personal emotional branch (PEB) and the contextual emotional branch (CEB). First, PEB is designed for the extraction of facial expression features and body posture features from the agent. A vision transformer (ViT), pretrained by contrastive learning with a novel loss function combining Euclidean distance and cosine similarity, is applied to enhance the robustness of facial expression features. Meanwhile, the human body contour images extracted by semantic segmentation are fed into another ViT to extract body posture features. Second, CEB is constructed for the extraction of global contextual features and interactive relationships among objects in the scene. The images masked by the agents' bodies are fed into a ViT to extract global contextual features. By leveraging both the gaze angle and depth map, a three-dimensional view graph (3DVG) is constructed to represent the interactive relationships between agents and objects in the scene. Then, a graph convolutional network is employed to extract interactive relationship features from the 3DVG. Finally, the multiplicative fusion strategy is applied to fuse the features of two branches, and the fused features are utilized to classify the emotions. TDRCer achieves an accuracy of 89.90% on the CAER-S dataset and a mean average precision (mAP) of 36.02% on the EMOTIons in context (EMOTIC) dataset. The code can be accessed at https://github.com/mengTender/TDRCer.

Human Action Recognition Based on Skeleton Information and Multi-Feature Fusion

Action assessment and feedback can effectively assist fitness practitioners in improving exercise benefits. In this paper, we address key challenges in human action recognition and assessment by proposing innovative methods that enhance performance while reducing computational complexity. Firstly, we present Oct-MobileNet, a lightweight backbone network, to overcome the limitations of the traditional OpenPose algorithm’s VGG19 network, which exhibits a large parameter size and high device requirements. Oct-MobileNet employs octave convolution and attention mechanisms to improve the extraction of high-frequency features from the human body contour, resulting in enhanced accuracy with reduced model computational burden. Furthermore, we introduce a novel approach for action recognition that combines skeleton-based information and multiple feature fusion. By extracting spatial geometric and temporal characteristics from actions, we employ a sliding window algorithm to integrate these features. Experimental results show the effectiveness of our approach, demonstrating its ability to accurately recognize and classify various human actions. Additionally, we address the evaluation of traditional fitness exercises, specifically focusing on the BaDunJin movements. We propose a multimodal information-based assessment method that combines pose detection and keypoint analysis. Label sequences are obtained through a pose detector and each frame’s keypoint coordinates are represented as pose vectors. Leveraging multimodal information, including label sequences and pose vectors, we explore action similarity and perform quantitative evaluations to help exercisers assess the quality of their exercise performance.

Automatic recognition of the neck–shoulder shape based on 2D photos

To improve body type classification and automatic identification, this study adopted a method by combining three-dimensional (3D) point-cloud data with two-dimensional (2D) photos for the classification of young mens’ neck–shoulder shape. A total of 180 male college students were measured by photo and 3D body measuring methods to get the main body information. As for the 3D body cloud data, reverse engineering software was used to measure the parameters closely related to the neck–shoulder shape. Five important parameters, including the forward angle, back angle, shoulder angle, the ratio between neck and shoulder width, and the ratio between neck width and thickness were analyzed to divide young men’s neck–shoulder shape into three categories, namely round drop shoulder, forward round neck and wide straight neck, and the classification rules were established. As for the 2D photos (such as front and side photos), the human body contour was extracted to obtain the parameters required for neck–shoulder shape classification by identifying the feature points. The neck–shoulder shape could be automatically recognized, and the accuracy rate is 97.7% to verify that the neck–shoulder shape recognition system is effective. This study proposed a simple neck–shoulder shape automatic recognition method based on the 2D body photos, and provide reference for automatic neck–shoulder shape recognition, 3D modeling and virtual fitting of local features.

Nominal Body Contour Reconstruction for Millimeter-Wave Characterization of Suicide Bomber Explosives

In order to improve the speed of passenger screening while preserving the effective capability to detect more sophisticated threats, airport security imaging systems must be able to accurately characterize concealed body-worn objects. In addition to improving the passenger experience, this system capability will enhance airport security for the traveling public. This article presents a real-time, fully automatic algorithm for the wideband millimeter-wave (mm-wave) radar reconstruction of the nominal human body contours, even in the presence of an affixed weak dielectric object or when a portion of the body cross section is not captured by the imaging scanner. The algorithm extracts the main contours from a noisy collection of 3-D reconstructed reflectivity and approximates the nominal human body cross section via fitting a low-order angular Fourier series. This important step is essential for precise characterization of concealed body-worn explosives. A ranking algorithm is developed as a metric for the nominal body reconstruction accuracy. We verify the developed algorithm by applying it to the actual images of the high definition–advanced imaging technology (HD-AIT) system, a laboratory prototype mm-wave scanning system developed recently by the U.S. Department of Homeland Security (DHS). The reconstructed body contours may be used to estimate the electric permittivity of the concealed person-worn objects.

DeepProfile: Accurate Under-the-Clothes Body Profile Estimation

Accurate human body profiles have many potential applications. Image-based human body profile estimation can be regarded as a fine-grained semantic segmentation problem, which is typically used to locate objects and boundaries in images. However, existing image segmentation methods, such as human parsing, require significant amounts of annotation and their datasets consider clothes as part of the human body profile. Therefore, the results they generate are not accurate when the human subject is dressed in loose-fitting clothing. In this paper, we created and labeled an under-the-clothes human body contour keypoint dataset; we utilized a convolutional neural network (CNN) to extract the contour keypoints, then combined them with a body profile database to generate under-the-clothes profiles. In order to improve the precision of keypoint detection, we propose a short-skip multi-scale dense (SMSD) block in the CNN to keep the details of the image and increase the information flow among different layers. Extensive experiments were conducted to show the effectiveness of our method. We demonstrate that our method achieved better results—especially when the person was dressed in loose-fitting clothes—than and competitive quantitative performance compared to state-of-the-art methods, while requiring less annotation effort. We also extended our method to the applications of 3D human model reconstruction and body size measurement.

Is skin pressure in load carriage over-evaluated?

Skin pressure is a biomechanical measure widely used in the assessment of load carriage systems. However, because of the complicated contour of human body, the stiffness of the pressure sensor array, and the large range of measurable pressures, there is much variability in previously reported results. In this paper, a simple mechanical model for load carriage was proposed, and the skin pressures beneath the shoulder and hip straps were predicted from the strap forces based on Laplace’s law. The proposed model was used to analyze data from literature with an aim to check the reliability of existing pressure measurements. The static and dynamic pressures at five locations on eight subjects wearing a backpack with a 10 kg load, while standing and walking on a treadmill, were measured respectively using pressure sensors of the air pack type. The combination of literature data analysis and experimental testing proved that the existing measurement method of interface pressure in load carriage systems often leads to over-estimation and this might misguide the pressure criteria set for load carriage system usage and design. The proposed model will be useful for quick prediction of the interface pressure in load carriage systems.

Accurate Human Parsing Model by Edge Contour and Pose Feature

<p indent=0mm>To address the problem of low accuracy and precision of human parsing due to human pose, edge contour, complexity of clothing accessories and occlusions of human pose joints in dressing scenes, an accurate human parsing model with edge contour and pose features is proposed for dressed human. Firstly, the backbone network based on ResNet-101 is used to represent input human body images and extract the coarse parsing features. Secondly, the edge contour module combining the global and local features after upsampling is constructed to obtain the edge contour of human body. Then, the defined human pose loss function based on human pose is added into pose estimation module to acquire pose features. Finally, coarse parsing feature, edge contour and pose features are integrated into accuracy parsing module, and the accurate human parsing results are output by the combined function of structure loss function and human parsing loss function. The experimental results show that the proposed model can effectively improve 1.96% of mIoU and accuracy on human datasets with more accurate segmentation results for different poses and occlusions of the human body parts.

Automatic recognition and 3D modeling of the neck-shoulder human shape based on 2D images

PurposeThis study focused on how to realize automatic recognition of young women's neck-shoulder shape based on the front and side images.Design/methodology/approachThe reverse engineering software was used to measure the body sizes of the neck-shoulder part based on the young women's three-dimensional (3D) point cloud data, and the important parameters closely related to the neck-shoulder shape were determined. The neck-shoulder shape of the subjects was classified to establish the classification rules. Then, based on the front and side images, the human body contour was extracted by Matlab, and the data required for neck-shoulder shape classification were obtained by identifying the feature points.FindingsThrough the cluster analysis based on the shoulder angle, back angle, shoulder depth/width ratio and armpit depth/width ratio, young women's neck-shoulder shape was divided into four categories, namely round wide shoulder, flat narrow shoulder, round drop shoulder and hunchback flat shoulder. The neck-shoulder shape could be automatically recognized based on the established classification rules and two-dimensional (2D) body measurement method, with an accuracy rate of 90%. The neck-shoulder shape automatic recognition system constructed based on this method is effective.Originality/valueThis study proposed a simple neck-shoulder automatic recognition method based on the 2D body images. This approach can be extended to other group of human body or other parts of the body.

A Gait Recognition Method Based on the Combination of Human Body Posture and Human Body Contour

Aiming at the issues that affect the gait recognition, such as outfit changes and carry-on-objects during gait recognition, this paper proposes a gait recognition method based on the improved GaitSet network, which relies on the fusion of human posture and human contour. This method introduces key points of human posture, and improves the precision of human contour extraction by introducing the key points of human posture. At the same time, the human posture map composed of the key points of human posture is used as the synchronous attribute of the contour map to extract gait features. Because key points focus on the inherent walking characteristics of human body, they are not affected by the external information such as clothing and carrying objects. Combined with the rich gait change attributes of human contour, it can effectively improve the accuracy and robustness of the gait recognition model. In scenes of human body wearing coats, our experimental results show that the accuracy of the proposed method in CASIA-B gait data set and self-built database can reach over 73.5% accuracy.

People Counting in Elevator Car Based on Computer Vision

To detect the number of passengers in an elevator based on computer vision technology, image preprocessing and human contour detection algorithms were proposed. After the target image was edge-detected and morphologically eroded, a binary image containing only human targets was obtained. The edges of the image were detected using canny operator, it was most suitable for the edge detection of the human body contour in elevator. Person counting was mainly performed by using morphological mark outline connection area counting method. According to the algorithm flow chart, the real elevator monitoring video was tested and verified. The number of people with fewer numbers (less than 10 persons) was highly accurate.

Spatiotemporal features of human motion for gait recognition

Gait is a novel biometric feature that offers human identification at a distance and without physical interaction with the imaging device. Moreover, it performs well even in low resolution which makes it ideal for use in numerous human identification applications, e.g.,visual surveillance, monitoring and access control systems. Most existing gait-based human identification solutions extract human body silhouettes, contours or shapes from the images and construct gait features. Therefore, the performance of such algorithms highly depends upon the accuracy of human body segmentation, which is still a challenging problem in the literature. In this paper, we propose a new gait recognition algorithm which uses the spatial and temporal motion characteristics of human gait for individual identification without needing the silhouette extraction. The proposed algorithm extracts a set of spatiotemporal local descriptors from the gait video sequences. The extracted descriptors are encoded using the Fisher vector encoding and Gaussian mixture model-based codebook. The encoded features are classified using a simple linear support vector machine to recognize the individuals. The proposed gait recognition method is evaluated on five widely used gait databases, including indoor (CMU MoBo, CASIA-B) and outdoor (NLPR, CASIA-C, TUM GAID) gait databases. The results reveal that our method showed excellent performance on all five databases and outperformed the state-of-the-art gait recognition approaches.

Template matching approach for automatic human body tracking in video

In this paper a novel template matching approach is presented to achieve automatic human body tracking in video sequences. The developed method which is based on a special template matching algorithm applied on a set of interest points detected on the human body contour. The matching approach is based on different types of similarity measures applied on consecutive frames from videos. Each frame was attacked with different types of noise: luminosity variation and motion blur. This new approach considers different matching constraints such as: cross-matching, uniqueness constraint and interest point's appearances and disappearances between consecutive frames. The algorithm was validated on two different datasets and the obtained results are so encouraging with high values of matching rate and good tracking rate.

High-Efficient Energy Harvester With Flexible Solar Panel for a Wearable Sensor Device

This paper proposes an optimal energy harvester (OEH) that uses a flexible photovoltaic (FPV) module to prolong battery life for a wearable body sensor node under indoor and outdoor conditions. The proposed sensor node uses a Bluetooth low-energy module, which consumes low power to wirelessly communicate with mobile devices for monitoring vital signals. A 19 cm $\times4$ cm FPV module that can easily conform to human body contours is used for generating a peak-power of up to 500 mW to charge the battery in outdoor environments. To optimize the power collected under various irradiances, the OEH is designed with a boost circuit for harvesting the low energy available indoors and a maximum power point tracker for collecting the high energy that is available outdoors. The booster can operate within a wide input voltage range of 0.65–3 V to generate a stable output voltage of 3.3 V, and it has a high conversion efficiency of ~95%. The MPPT uses fuzzy inference engine to find the MPP of the FPV that is required to power the system. The results of our experiment show that the proposed device can prolong the lifetime by up to 950% under an irradiance of only 2.9 mW/cm2, and the proposed node becomes an autonomous node under partial shading and sunlight conditions. Therefore, this paper can provide a useful solution for extending the lifetime of wearable sensor devices.

Study on Human Body Contour Extraction from Images Based on HSV Color Model

Study on Human Body Contour Extraction from Images Based on HSV Color Model

人の輪郭と局所特徴量を用いた立礼教育システムにおける屈体角度の推定

人の輪郭と局所特徴量を用いた立礼教育システムにおける屈体角度の推定

Locally adaptive combining colour and depth for human body contour tracking using level set method

In this study, the authors present a novel human body contour tracking method which adaptively combines colour and depth cues of RGB‐D images in level set framework. The authors model the body object by active contour. When the body object is far away from the background objects, it is relatively easy to separate the object from the background using the depth cue. In this case, the depth cue should dominate the evolution of the active contour. When some part of the body is very close to the background objects, the discriminability of the depth decreases rapidly. In this local region the colour cue should dominate the evolution, whereas the depth cue should still play an important role in other regions. To achieve these objectives, the authors propose to use a superpixel‐based locally adaptive weight map to determine the importance of the depth cue. Moreover, to obtain more accurate contours and to avoid error drifting, based on the two novel properties of the human body surface in the depth images, the authors propose two simple but effective algorithms to refine the tracking results of the level set method. The promising results demonstrate the performance of the proposed method.

Autonomous Wearable Sensor Nodes With Flexible Energy Harvesting

Distributed wearable wireless sensors are widely employed in wireless body sensor network for various physiological monitoring purposes like health or performance related monitoring applications. The real challenges in employing these wearable wireless sensors on human subjects include: 1) bulky and rigid system design thus, it is difficult to conform to human body contour and 2) limited operational lifespan of batteries with finite energy supply. In this paper, an autonomous body-worn wireless sensor node with flexible energy harvesting (FEH) mechanism, able to conform to body contour, is proposed for biometric monitoring. To be totally sustainable and compact, the FEH mechanism is equipped with an ultralow power management circuit (PMC) specially designed on a flexible PCB. The flexible PMC is able to transfer near maximum electrical power from the input solar energy source to store in the supercapacitor for powering the wireless sensor node. The power consumption of the flexible PMC is only 32.86 $\mu{\rm W}$ . Experimental results show that under indoor condition, typical average lighting intensity of 320 lux, the wearable sensor node is able to continuously monitor the temperature of the wearer, read, and transmit back to the base node in a wireless manner, without the need of any battery. In addition, the designed FEH sensor node flexed onto the wearer body contour at an angle of 30 $^{\circ}$ generates 56 $\mu{\rm W}$ of electrical power, sufficient to sustain its operation for ${>}{\rm 15}~{\rm h}$ .

Application of Color Transfer Algorithm in the Virtual Color Restoration of Ancient Architecture

An automatic human motion tracking algorithm based the connection between the improved Snake model and optical flow arithmetic is proposed. The initial contour most close to human body is acquired by corner detection arithmetic, which decreases the iterations and reduces the probability of converging at local extreme value of Snake model. Because the tracking with Snake model is unstable and often loses the object, the good feature points are chosen from the contour points in current frame for optical flow estimation. Then the result is chosen as the initial contour in next frame. Experimental results show that improved Snake model can be deformed to the actual human body contour, and the automatic and real-time human tracking is implemented.

Cascade Classifier Using Combination of Histograms of Oriented Gradients for Rapid Pedestrian Detection

Accurate and efficient human detection has become an important area for research in computer vision. In order to solve problems in the past human detection algorithms such as features with fixed sizes, fixed positions and fixed number, the human detection based on united Hogs algorithm was proposed. This algorithm can dynamically generate the features closer to human body contours. Basically maintaining the detection speed, the detection accuracy was improved by our algorithm. We demonstrate that comparing with human detection algorithm using haar, traditional hogs and hogs with variable-size blocks, our algorithm is better in both of detection rate and false positive rate.

Three experiments to support the design of lightweight comfortable vehicle seats

Seats need to be more lightweight for airplanes, cars, busses and even trains to contribute to a better environment and to reduce energy consumption. However, a reduction in comfort due to weight reduction is not preferable, which opens a new area of research: improving comfort with a minimum of material or with lightweight materials and systems. In this paper three experiments are performed to test the effects of light weight seats and parts of a seat on comfort. The first experiment shows that a new developed light weight massage system improves comfort and reduces muscle activity. The second experiment shows that the automatic seat adjustment without motors improves the comfort as well. The third experiment showed that a light weight seat following closely the human body contour is experienced on many aspects in the same way as current more heavy seats. More research and models will be needed in this ergonomic field which needs more attention.