Anna Vögele

Efficient Unsupervised Temporal Segmentation of Motion Data

We introduce a method for automated temporal segmentation of human motion data into distinct actions and compositing motion primitives based on self-similar structures in the motion sequence. We use neighborhood graphs for the partitioning and the similarity information in the graph is further exploited to cluster the motion primitives into larger entities of semantic significance. The method requires no assumptions about the motion sequences at hand and no user interaction is required for the segmentation or clustering. In addition, we introduce a feature bundling preprocessing technique to make the segmentation more robust to noise, as well as a notion of motion symmetry for more refined primitive detection. We test our method on several sensor modalities, including markered and markerless motion capture as well as on electromyograph and accelerometer recordings. The results highlight our systems capabilities for both segmentation and for analysis of the finer structures of motion data, all in a completely unsupervised manner.

One Small Step for a Man: Estimation of Gender, Age and Height from Recordings of One Step by a Single Inertial Sensor

A number of previous works have shown that information about a subject is encoded in sparse kinematic information, such as the one revealed by so-called point light walkers. With the work at hand, we extend these results to classifications of soft biometrics from inertial sensor recordings at a single body location from a single step. We recorded accelerations and angular velocities of 26 subjects using integrated measurement units (IMUs) attached at four locations (chest, lower back, right wrist and left ankle) when performing standardized gait tasks. The collected data were segmented into individual walking steps. We trained random forest classifiers in order to estimate soft biometrics (gender, age and height). We applied two different validation methods to the process, 10-fold cross-validation and subject-wise cross-validation. For all three classification tasks, we achieve high accuracy values for all four sensor locations. From these results, we can conclude that the data of a single walking step (6D: accelerations and angular velocities) allow for a robust estimation of the gender, height and age of a person.

Interactive steering of mesh animations

Creating geometrically detailed mesh animations is an involved and resource-intensive process in digital content creation. In this work we present a method to rapidly combine available sparse motion capture data with existing mesh sequences to produce a large variety of new animations. The key idea is to model shape changes correlated to the pose of the animated object via a part-based statistical shape model. We observe that compact linear models suffice for a segmentation into nearly rigid parts. The same segmentation further guides the parameterization of the pose which is learned in conjunction with the marker movement. Besides the inherent high geometric detail, further benefits of the presented method arise from its robustness against errors in segmentation and pose parameterization. Due to efficiency of both learning and synthesis phase, our model allows to interactively steer virtual avatars based on few markers extracted from video data or input devices like the Kinect sensor.

Novel Methods for Surface EMG Analysis and Exploration Based on Multi-Modal Gaussian Mixture Models

This paper introduces a new method for data analysis of animal muscle activation during locomotion. It is based on fitting Gaussian mixture models (GMMs) to surface EMG data (sEMG). This approach enables researchers/users to isolate parts of the overall muscle activation within locomotion EMG data. Furthermore, it provides new opportunities for analysis and exploration of sEMG data by using the resulting Gaussian modes as atomic building blocks for a hierarchical clustering. In our experiments, composite peak models representing the general activation pattern per sensor location (one sensor on the long back muscle, three sensors on the gluteus muscle on each body side) were identified per individual for all 14 horses during walk and trot in the present study. Hereby we show the applicability of the method to identify composite peak models, which describe activation of different muscles throughout cycles of locomotion.

FuryExplorer: visual-interactive exploration of horse motion capture data

The analysis of equine motion has a long tradition in the past of mankind. Equine biomechanics aims at detecting characteristics of horses indicative of good performance. Especially, veterinary medicine gait analysis plays an important role in diagnostics and in the emerging research of long-term effects of athletic exercises. More recently, the incorporation of motion capture technology contributed to an easier and faster analysis, with a trend from mere observation of horses towards the analysis of multivariate time-oriented data. However, due to the novelty of this topic being raised within an interdisciplinary context, there is yet a lack of visual-interactive interfaces to facilitate time series data analysis and information discourse for the veterinary and biomechanics communities. In this design study, we bring visual analytics technology into the respective domains, which, to our best knowledge, was never approached before. Based on requirements developed in the domain characterization phase, we present a visual-interactive system for the exploration of horse motion data. The system provides multiple views which enable domain experts to explore frequent poses and motions, but also to drill down to interesting subsets, possibly containing unexpected patterns. We show the applicability of the system in two exploratory use cases, one on the comparison of different gait motions, and one on the analysis of lameness recovery. Finally, we present the results of a summative user study conducted in the environment of the domain experts. The overall outcome was a significant improvement in effectiveness and efficiency in the analytical workflow of the domain experts.

Visual-interactive segmentation of multivariate time series

Choosing appropriate time series segmentation algorithms and relevant parameter values is a challenging problem. In order to choose meaningful candidates it is important that different segmentation results are comparable. We propose a Visual Analytics (VA) approach to address these challenges in the scope of human motion capture data, a special type of multivariate time series data. In our prototype, users can interactively select from a rich set of segmentation algorithm candidates. In an overview visualization, the results of these segmentations can be compared and adjusted with regard to visualizations of raw data. A similarity-preserving colormap further facilitates visual comparison and labeling of segments. We present our prototype and demonstrate how it can ease the choice of winning candidates from a set of results for the segmentation of human motion capture data.

Automatic Temporal Segmentation of Articulated Hand Motion

This paper introduces a novel and efficient segmentation method designed for articulated hand motion. The method is based on a graph representation of temporal structures in human hand-object interaction. Along with the method for temporal segmentation we provide an extensive new database of hand motions. The experiments performed on this dataset show that our method is capable of a fully automatic hand motion segmentation which largely coincides with human user annotations.

Sleep Detection Using a Depth Camera

The work at hand presents a method to assess the quality of human sleep within a non-laboratory environment. The monitoring of patients is performed by means of a Kinect device. This results in a non-invasive method which is independent of immediate physical contact to subjects. The results of a study which was carried out as proof of concept are discussed and compared with the polysomnography-based gold standard of sleep analysis.

Approaches and Challenges in the Visual-interactive Comparison of Human Motion Data.

Many analysis goals involving human motion capture (MoCap) data require the comparison of motion patterns. Pioneer works in visual analytics recently recognized visual comparison as substantial for visual-interactive analysis. This work reflects the design space for visual-interactive systems facilitating the visual comparison of human MoCap data, and presents a taxonomy comprising three primary factors, following the general visual analytics process: algorithmic models, visualizations for motion comparison, and back propagation of user feedback. Based on a literature review, relevant visual comparison approaches are discussed. We outline remaining challenges and inspiring works on MoCap data, information visualization, and visual analytics.