Dynamics Of Human Motion Research Articles

Modelling and predicting forced migration.

Migration models have evolved significantly during the last decade, most notably the so-called flow Fixed-Effects (FE) gravity models. Such models attempt to infer how human mobility may be driven by changing economy, geopolitics, and the environment among other things. They are also increasingly used for migration projections and forecasts. However, recent research shows that this class of models can neither explain, nor predict the temporal dynamics of human movement. This shortcoming is even more apparent in the context of forced migration, in which the processes and drivers tend to be heterogeneous and complex. In this article, we derived a Flow-Specific Temporal Gravity (FTG) model which, compared to the FE models, is theoretically similar (informed by the random utility framework), but empirically less restrictive. Using EUROSTAT data with climate, economic, and conflict indicators, we trained both models and compared their performances. The results suggest that the predictive power of these models is highly dependent on the length of training data. Specifically, as time-series migration data lengthens, FTG's predictions can be increasingly accurate, whereas the FE model becomes less predictive.

Effect of trained evacuation leaders on victims’ safety during an active shooter incident

Trained evacuation leaders in emergency offer the potential for improved decision making and evacuation. Compared to victims, trained evacuation leaders can make educated assessments of the situation based on their training, knowledge of the facilities, and additional details about the incident, which enables them to guide victims in choosing a safe departure time and evacuation route. Despite a general understanding about the benefits of such leaders in evacuation, mass shooting cases require a separate attention because these cases are more complex with different behavioral decisions, not just running away, with a continuously changing source of the hazard source, the shooter. This study develops a simulation model package and evaluates the effect of trained evaluation leaders on the victim safety during an active shooter incident. The study leverages sophisticated human motion dynamics models and human behaviors supported by past literature in an agent-based model. The study varies several parameters (e.g., occupancy, firing rates and gun range, and victims’ decision of running or hiding) in this simulation to draw generalized conclusions on the leaders’ impact on various scenarios. The results reveal general findings with several interesting points. Overall, increased leaders’ presence contributes to fewer fatalities. Even few trained leaders, compared with none, can considerably improve victim safety. Even if leaders are not uniformly positioned, they still provide substantial benefits for victim’s safety. The leaders’ benefits were consistently found in various parametric studies (e.g., number of leaders, occupancy, leaders’ strategic placement, gun range, and shooting rate) that support the mentioned findings.

Fabrication and Experimental Validation of a Sensitive and Robust Tactile Sensing Array with a Micro-Structured Porous Dielectric Layer.

The development of pressure sensors of high sensitivity and stable robustness over a broad range is indispensable for the future progress of electronic skin applicable to the detection of normal and shear pressures of various dynamic human motions. Herein, we present a flexible capacitive tactile sensing array that incorporates a porous dielectric layer with micro-patterned structures on the surface to enable the sensitive detection of normal and shear pressures. The proposed sensing array showed great pressure-sensing performance in the experiments, with a broad sensing range from several kPa to 150 kPa of normal pressure and 20 kPa of shear pressure. Sensitivities of 0.54%/kPa at 10 kPa and below, 0.45%/kPa between 10 kPa and 80 kPa, and 0.12%/kPa at 80 kPa and above were achieved for normal pressures. Meanwhile, for shear pressures, sensitivities up to 1.14%/kPa and 1.08%/kPa in x and y directions, respectively, and below 10 kPa, 0.73%/kPa, and 0.75%/kPa under shear pressure over 10 kPa were also validated. The performance of the finger-attached sensing array was also demonstrated, demonstrating which was a potential electronic skin to use in all kinds of wearable devices, including prosthetic hands, surgical robots, and other pressure monitoring systems.

Spatial–temporal modeling for prediction of stylized human motion

Human motion prediction refers to forecasting human motion in the future given a past motion sequence, which has significant applications in human tracking, automatic motion generation, autonomous driving, human-robotics interaction, etc. Previous works usually used RNN-based methods, focusing on modeling the temporal dynamics of human motion, which have made great effort on content motions. However, it is unclear for their performance on stylized motion, which is with more expressive emotions and states of the human motion. Different styles within the same motion type have similar motion patterns but also subtle variances. This makes it difficult to be predicted. The main idea of this paper is to learn the spatial characteristic of stylized motion and combine it with the temporal dynamics to achieve accurate prediction. We adopt a transformer-based style encoder to learn the motion representation in the pose space and then maps it to the latent space modeled by the constant variance Gaussian mixture model; meanwhile, we use the hierarchical multi-scale RNN as a temporal encoder to capture the temporal dynamics of human motion; finally, we feed the spatial and temporal features into the prediction decoder to predict the next frame. Our experiments on the Human 3.6 M and Stylized MotionDatasets demonstrate that our model has comparable prediction performance with the state-of-the-art motion prediction works on Human 3.6 M and outperforms previous works on stylized human motion prediction.

Comparison of several muscle modeling alternatives for computationally intensive algorithms in human motion dynamics

Several approaches are currently employed to address the predictive simulation of human motion, having in common their high computational demand. Muscle modeling seems to be an essential ingredient to provide human likeness to the obtained movements, at least for some activities, but it increases even more the computational load. This paper studies the efficiency and accuracy yielded by several alternatives of muscle modeling in the forward-dynamics analysis of captured motions, as a method that encompasses the computationally intensive character of predictive simulation algorithms with a known resulting motion which simplifies the comparisons. Four muscle models, the number of muscles, muscle torque generators, muscular synergies, and look-up tables for musculotendon lengths and moment arms are considered and analyzed, seeking to provide criteria on how to include the muscular component in human multibody models so that its effect on the resulting motion is captured while keeping a reasonable computational cost. Gait and vertical jump are considered as examples of slow- and fast-dynamics motions. Results suggest that: (i) the rigid-tendon model with activation dynamics offers a good balance between accuracy and efficiency, especially for short-tendon muscles; (ii) including muscles in the model leads to a decrease in efficiency which is highly dependent on the muscle model employed and the number of muscles considered; (iii) muscle torque generators keep the efficiency of skeletal models; (iv) muscular synergies offer almost no advantage for this problem; and (v) look-up tables for configuration-dependent kinematic magnitudes have a non-negligible impact on the efficiency, especially for simplified muscle models.

Gesture Tracking and Recognition Algorithm for Dynamic Human Motion Using Multimodal Deep Learning

To address the problems of the traditional human motion gesture tracking and recognition methods, such as poor tracking effect, low recognition accuracy, high frame loss rate, and long-time cost, a dynamic human motion gesture tracking and recognition algorithm using multimode deep learning was proposed. Firstly, the collected human motion images are repaired in the three-dimensional (3D) environment, and the multimodal 3D human motion model is reconstructed using the processed images. Secondly, according to the results of model reconstruction, the camera gesture and other parameters of the keyframe are used to construct the target tracking optimization function so as to achieve the purpose of accurate tracking of human motion. Finally, for multimodal human motion gesture learning, a convolutional neural network (CNN) is developed. The trained CNN is utilized to complete dynamic human motion recognition after convolutional and pooling calculations. The results demonstrate that the proposed algorithm is effective in tracking human motion gestures. The average recognition accuracy is 96%, the average frame loss rate is 8.8%, the time cost is low, and the proposed algorithm has a high F-measure and much lower power consumption than other algorithms, indicating that the proposed algorithm is effective.

Is coordination variability using vector coding different in overground and treadmill walking and running?

BackgroundCoordination variability has been linked to overuse running injuries and has been studied both on a treadmill and over-ground. It is not clear, however, if the coordination variability data from over-ground locomotion can be compared with treadmill locomotion data. Research questionTherefore, the purpose of this study was to compare coordination variability of selected lower extremity couplings at different locomotor speeds during over-ground and treadmill walking and running. MethodsNineteen (10 female, 9 male) healthy, recreational collegiate runners participated in this study. Each participant performed in two different conditions: over-ground and on a treadmill at three walking speeds (1.2, 1.6, and 2.0 m•s−1) and three running speeds (2.8, 3.2, and 3.6 m•s−1). A modified vector coding technique was used to calculate coordination variability for five selected coupled segment and joint angles. Each of the segmental couples was analyzed separately using a two-way repeated measures ANOVA (Condition Χ Speed) implemented with one-dimensional statistical parametric mapping. ResultsWhile no interaction effects were observed for condition X speed, we saw increased coordination variability in the sagittal couples during overground compared with treadmill locomotion, which predominantly occurred during the stance phase. There were mixed results for changes in coordination variability as a function of gait speed. However, for the sagital plane couplings, coordination variability decreased with speed, particularly during the stance phase. SignificanceThese results suggest that the controlled belt speed of the treadmill affects the intrinsic dynamics of human movement and this should be considered when making comparisons between treadmill and over-ground studies and in future study designs.

Dual-Stream Structured Graph Convolution Network for Skeleton-Based Action Recognition

In this work, we propose a dual-stream structured graph convolution network ( DS-SGCN ) to solve the skeleton-based action recognition problem. The spatio-temporal coordinates and appearance contexts of the skeletal joints are jointly integrated into the graph convolution learning process on both the video and skeleton modalities. To effectively represent the skeletal graph of discrete joints, we create a structured graph convolution module specifically designed to encode partitioned body parts along with their dynamic interactions in the spatio-temporal sequence. In more detail, we build a set of structured intra-part graphs, each of which can be adopted to represent a distinctive body part (e.g., left arm, right leg, head). The inter-part graph is then constructed to model the dynamic interactions across different body parts; here each node corresponds to an intra-part graph built above, while an edge between two nodes is used to express these internal relationships of human movement. We implement the graph convolution learning on both intra- and inter-part graphs in order to obtain the inherent characteristics and dynamic interactions, respectively, of human action. After integrating the intra- and inter-levels of spatial context/coordinate cues, a convolution filtering process is conducted on time slices to capture these temporal dynamics of human motion. Finally, we fuse two streams of graph convolution responses in order to predict the category information of human action in an end-to-end fashion. Comprehensive experiments on five single/multi-modal benchmark datasets (including NTU RGB+D 60, NTU RGB+D 120, MSR-Daily 3D, N-UCLA, and HDM05) demonstrate that the proposed DS-SGCN framework achieves encouraging performance on the skeleton-based action recognition task.

A study of human motion in computer vision systems based on a skeletal model

Methods of studying human motion in computer vision systems can be divided into two types. These are analysis in two-dimensional and three-dimensional space. The former uses a single camera image and/ or multiple body sensors. Such an approach leads to a rapid accumulation of error and, consequently, low accuracy of the figure representation. Multiple cameras are usually used in the case of three-dimensional space analysis, while the objects are represented as sets of volumetric elements. Despite the high accuracy of this method, it is associated with high computational complexity and internal network load. The purpose of the paper is to develop a model using a single camera, while approaching three-dimensional space analysis methods in terms of accuracy. In this paper a human figure is represented as a skeleton. The skeleton is described by an acyclic connected graph. The general structure of a human figure is analyzed. Fifteen basic points are selected. Physical and logical connections between them were studied and mathematically described. The velocity and spatial characteristics of the points and connections outline the general dynamics of motion. The study describes a model of human motion and gives the option for model construction on the example of a particular image. The developed algorithm for collection and analysis of information estimates relative locations and velocity characteristics of the graph elements. The model can be used for acquisition of information about the reference dynamics of human movements. In case of detecting major differences between the reference and the reality, the behavior is defined as deviant. Thus, the obtained algorithm can be applied in computer vision systems for detection and analysis of human movements.

CoolMoves

Current Virtual Reality (VR) systems are bereft of stylization and embellishment of the user's motion - concepts that have been well explored in animations for games and movies. We present CooIMoves, a system for expressive and accentuated full-body motion synthesis of a user's virtual avatar in real-time, from the limited input cues afforded by current consumer-grade VR systems, specifically headset and hand positions. We make use of existing motion capture databases as a template motion repository to draw from. We match similar spatio-temporal motions present in the database and then interpolate between them using a weighted distance metric. Joint prediction probability is then used to temporally smooth the synthesized motion, using human motion dynamics as a prior. This allows our system to work well even with very sparse motion databases (e.g., with only 3-5 motions per action). We validate our system with four experiments: a technical evaluation of our quantitative pose reconstruction and three additional user studies to evaluate the motion quality, embodiment and agency.

Understanding collective human movement dynamics during large-scale events using big geosocial data analytics

Conventional approaches for modeling human mobility pattern often focus on human activity and movement dynamics in their regular daily lives and cannot capture changes in human movement dynamics in response to large-scale events. With the rapid advancement of information and communication technologies, many researchers have adopted alternative data sources (e.g., cell phone records, GPS trajectory data) from private data vendors to study human movement dynamics in response to large-scale natural or societal events. Big geosocial data such as georeferenced tweets are publicly available and dynamically evolving as real-world events are happening, making it more likely to capture the real-time sentiments and responses of populations. However, precisely-geolocated geosocial data is scarce and biased toward urban population centers. In this research, we developed a big geosocial data analytical framework for extracting human movement dynamics in response to large-scale events from publicly available georeferenced tweets. The framework includes a two-stage data collection module that collects data in a more targeted fashion in order to mitigate the data scarcity issue of georeferenced tweets; in addition, a variable bandwidth kernel density estimation(VB-KDE) approach was adopted to fuse georeference information at different spatial scales, further augmenting the signals of human movement dynamics contained in georeferenced tweets. To correct for the sampling bias of georeferenced tweets, we adjusted the number of tweets for different spatial units (e.g., county, state) by population. To demonstrate the performance of the proposed analytic framework, we chose an astronomical event that occurred nationwide across the United States, i.e., the 2017 Great American Eclipse, as an example event and studied the human movement dynamics in response to this event. However, this analytic framework can easily be applied to other types of large-scale events such as hurricanes or earthquakes.

Moving Forward: A Bioarchaeology of Mobility and Migration

Growing interest in bioarchaeology and its ability to address complex questions tied to social and biological identities in the past has led to the development of nuanced methods for evaluating mobility and migration using human skeletal remains. Improving our ability to identify both short- and long-term migration through observations of body modification, analyses of biological distance, and applications of biogeochemical and aDNA techniques has enabled us to move beyond the simple dichotomous classification of past individuals as either local or nonlocal. These approaches have elucidated the complexity of migration processes while also revealing the heterogeneous ways in which individual agents and social groups incorporate, instigate, experience, and adapt to movement. These data have likewise demonstrated the potential of bioarchaeology to reveal broader patterns of social organization, social and ethnic identities, fictive kinship, postmarital residence, gender roles and relations, detailed life courses, responses to climate stress, and pathways of disease transmission. As bioarchaeology continues to contribute to mobility and migration studies, human skeletal data should be further contextualized by the archaeological record and linked to anthropological, archaeological, and bioarchaeological theoretical frameworks as part of more holistic attempts to explain the diversity and dynamics of human movement, interaction, and identity construction among communities in the past.

Online Non-Collocated Estimation of Payload and Articular Stress for Real-Time Human Ergonomy Assessment

Improving the quality of work for human beings is receiving a lot of attention from multiple research communities. In particular, digital transformation in human factors and ergonomics is going to empower the next generation of the socio-technical workforce. The use of wearable sensors, collaborative robots, and exoskeletons, coupled with novel technologies for the real-time assessment of human ergonomy forms the crux of this digital transformation. In this direction, this paper focuses on the open problem of estimating the interaction wrench experienced at the human extremities (such as hands), where the feasibility of direct sensor measurements is not practical. We refer to our approach as non-collocated wrench estimation, as we aim to estimate the wrench at known contact locations but without using any direct force-torque sensor measurements at these known locations. We achieve this by extending the formulation of stochastic inverse dynamics for humans by considering a centroidal dynamics constraint to perform a reliable non-collocated estimation of interaction wrench and the joint torques (articular stress) experienced as a direct consequence of the interaction. Our approach of non-collocated estimation is thoroughly validated in terms of payload estimation and articular stress estimation through validation and experimental scenarios involving dynamic human motions like walking.

Tracing mobility patterns through the 6th-5th millennia BC in the Carpathian Basin with strontium and oxygen stable isotope analyses.

The complexity of Neolithic population movements and their interpretation through material culture have been the subject of archaeological research for decades. One of the dominant narratives proposes that groups from the Starčevo-Körös-Criş complex spread from the central towards the northern Balkans in the Early Neolithic and eventually brought the Neolithic lifestyle into present-day Hungary. Broad geographical migrations were considered to shape the continuous expansion of Neolithic groups and individuals. However, recent archaeological research, aDNA, and isotope analyses challenged the synchronous appearance of specific material culture distributions and human movement dynamics through emphasizing communication networks and socio-cultural transformation processes. This paper seeks to retrace the complexity of Neolithic mobility patterns across Hungary by means of strontium and oxygen stable isotope analyses, which were performed on a total of 718 human dental enamel samples from 55 Neolithic sites spanning the period from the Starčevo to the Balaton-Lasinja culture in Transdanubia and from the Körös to the Tiszapolgár cultural groups on the Great Hungarian Plain (Alföld). This study presents the largest strontium and oxygen isotope sample size for the Neolithic Carpathian Basin and discusses human mobility patterns on various geographical scales and throughout archaeological cultures, chronological periods, and sex and gender categories in a multiproxy analysis. Based on our results, we discuss the main stages of the Neolithisation processes and particularly trace individual movement behaviour such as exogamy patterns within extensive social networks. Furthermore, this paper presents an innovative differentiation between mobility patterns on small, micro-regional, and supra-regional scales, which provides new insights into the complex organisation of Neolithic communities.

Silicone-Ionic Liquid Elastomer Composite with Keratin as Reinforcing Agent Utilized as Pressure Sensor.

Development of a flexible pressure sensor is crucial for the future improvement of the wearable electronic devices designed to detect dynamic human motion. In this study, a novel pressure sensor with remarkably improved force sensing characteristics is obtained through combined usage of polydimethylsiloxane (PDMS) and ionic liquid (IL). Keratin is dispersed homogeneously in the PDMS matrix to serve as a reinforcing filler. High conductivity IL is employed as sensitivity-enhancing constituent in the elastomer, and the effect of the amount of IL on elastomers' pressure-sensing performance is investigated. The elastomer with 70 parts per hundred rubber (phr) IL shows excellent pressure-sensing performance. This novel pressure sensor demonstrates high linear sensitivity (0.037 kPa-1 ) in the large pressure region of 0-10kPa. Response and recovery times are 8 and 11ms, respectively, which are much shorter than previously reported. Moreover, the pressure sensor could distinguish different pressures via stable sensing signals in the pressure range of 0 to 50kPa. The excellent performance of the novel pressure sensor has application potential in various fields, such as health monitoring and soft robotics.

Microbiome Innovation in Agriculture: Development of Microbial Based Tools for Insect Pest Management

According to the Entomological Society of America’s (ESA) recent position statement, invasive insects incur control costs of over $2.5 billion and cause economic damages to crops, lawns, forests, and pastures totaling $18 billion per year (“The Not-So-Hidden Dangers of Invasive Species” 2018). The threat of invasive crop pests to food security continues to be driven by complex dynamics of human movements, global trade activities, climate change, and changing agricultural practices. Chemical insecticides have been central to crop protection against invasive insects. However, a growing demand for reduced agricultural chemical use due to environmental and human health concerns in addition to pesticide resistance issues are fueling interests in innovative approaches to manage invasive pests. For decades, the role of microbes in pest management has been largely confined to using entomopathogens, with only a handful of microbial species being developed into bioinsecticides. The paradigm is shifting owing to the advent of high-throughput sequencing, functional omics, and gene editing technologies, which greatly accelerate microbial discovery, as well as a better understanding of microbial functions in complex communities across different habitats. There is also overwhelming evidence that symbiotic microbes play pivotal roles in shaping various insect traits. The collection of microbes associated with a given environment (both biotic and abiotic) and their collective genetic materials is termed the microbiome. In this paper, we survey the latest development in microbiome research that could be leveraged into novel management tools for agricultural insect pests (Figure 1). We first review the diverse insect traits shaped by insect-microbe associations that span nutrition, immunity, ecological interactions with natural enemy, insecticide resistance, and behavior. This is followed by a discussion on different microbiome manipulation approaches to alter pest traits, describing some of the opportunities and obstacles for each approach. We then highlight microbiomes as untapped chemical inventories to discover novel biopesticides, including plant-incorporated protectants and semiochemicals. The last topic covered is the use of beneficial microbes to improve performance of mass-reared insects for autocidal programs, including sterile insect technique and incompatible insect technique, in which we identify topics where data are limited or inconclusive for future research.

A REAL-TIME PHOTOGRAMMETRIC SYSTEM FOR MONITORING HUMAN MOVEMENT DYNAMICS

Abstract. The human body posture is rich with dynamic information that can be captured by algorithms, and many applications rely on this type of data (e.g., action recognition, people re-identification, human-computer interaction, industrial robotics). The recent development of smart cameras and affordable red-green-blue-depth (RGB-D) sensors has enabled cost-efficient estimation and tracking of human body posture. However, the reliability of single sensors is often insufficient due to occlusion problems, field-of-view limitations, and the limited measurement distances of the RGB-depth sensors. Furthermore, a large-scale real-time response is often required in certain applications, such as physical rehabilitation, where human actions must be detected and monitored over time, or in industries where human motion is monitored to maintain predictable movement flow in a shared workspace. Large-scale markerless motion-capture systems have therefore received extensive research attention in recent years.In this paper, we propose a real-time photogrammetric system that incorporates multithreading and a graphic process unit (GPU)-accelerated solution for extracting 3D human body dynamics in real-time. The system includes a stereo camera with preliminary calibration, from which left-view and right-view frames are loaded. Then, a dense image-matching algorithm is married with GPU acceleration to generate a real-time disparity map, which is further extended to a 3D map array obtained by photogrammetric processing based on the camera orientation parameters. The 3D body features are acquired from 2D body skeletons extracted from regional multi-person pose estimation (RMPE) and the corresponding 3D coordinates of each joint in the 3D map array. These 3D body features are then extracted and visualised in real-time by multithreading, from which human movement dynamics (e.g., moving speed, knee pressure angle) are derived. The results reveal that the process rate (pose frame-rate) can be 20 fps (frames per second) or above in our experiments (using two NVIDIA 2080Ti and two 12-core CPUs) depending on the GPU exploited by the detector, and the monitoring distance can reach 15 m with a geometric accuracy better than 1% of the distance.This real-time photogrammetric system is an effective real-time solution to monitor 3D human body dynamics. It uses low-cost RGB stereo cameras controlled by consumer GPU-enabled computers, and no other specialised hardware is required. This system has great potential for applications such as motion tracking, 3D body information extraction and human dynamics monitoring.

CNT-coated magnetic self-assembled elastomer micropillar arrays for sensing broad-range pressures

Cost-effective and broad-range flexible pressure sensors are highly desirable for use in wearable electronic devices for a wide range of sensitive measurements, from pulse to dynamic human motion. Herein, we present a high-performance flexible pressure sensor based on carbon nanotube-coated elastomer micropillar arrays (EMPAs). The EMPAs are prepared using a facile magnetic-field-induced self-assembly technique, which provides a simple and low-cost strategy to fabricate microstructured elastomers for constructing high-performance pressure sensors. Since high-ratio microstructured elastomers are employed as the sensing materials, our pressure sensors show a superior pressure-sensing performance, with a high sensitivity of 0.497 kPa−1 in the low-pressure regime of <100 Pa, a broad sensing range from a few pascals up to 80 kPa, and a fast response time of <0.2 s. Finally, we demonstrate that the sensor can be used to not only detect human arthrosis movements, but also to monitor subtle human physiological activities.

Human Oral Motion-Powered Smart Dental Implant (SDI) for In Situ Ambulatory Photo-biomodulation Therapy.

Peri-implant disease is an inflammatory condition affecting the soft and hard tissues surrounding a dental implant. However, current preventative methods are insufficient due to the limited bioactivity on the dental implant and poor patient compliance. Recently, photo-biomodulation (PBM) therapy that can recover and regenerate peri-implant soft tissue has attracted considerable attention in dentistry. In this paper, a seamless human oral motion-powered dental implant system (called Smart Dental Implant or SDI) is presented as an ambulatory PBM therapy modality. SDI allows the in situ light delivery, which is enabled by the energy harvesting from dynamic human oral motions (chewing and brushing) via an engineered piezoelectric dental crown, an associated circuit, and micro light emitting diodes (LEDs). The SDI also offers adequate mechanical strength as the clinical standards. Using primary human gingival keratinocytes (HGKs) as a model host organism and Pseudomonas aeruginosa lipopolysaccharides (LPS) as a model inflammatory stimulus, effective SDI-mediated PBM therapy is demonstrated. A new class of dental implants could be an ambulatory PBM therapy platform for the prevention of peri-implant disease without patient dependency, warranting long-lasting dental implants.

Dynamic analysis of stepping behavior of pedestrian social groups on stairs

Human movement dynamics is highly correlated with stepping locomotion. This paper aims to explore stepping behavior of pedestrian social groups on stairs through a field observation. A total of 96 singles and 194 pedestrian groups with different sizes were adopted as test subjects. Their stepping features such as speed, step length, step time and step width were obtained according to trajectory extraction and curvature calculation. Relations between these variables during ascending and descending movement on stairs are analyzed and compared. The effects of group sizes and movement processes on stepping locomotion are also discussed. It is discovered that both step time and step width for groups with larger size are significantly higher during descending movement. This study provides a new insight into stepping behavior of pedestrian groups on stairs, and is beneficial to group movement modelling and pedestrian facility design.