Chuanjun Li

Real-time classification of variable length multi-attribute motions

Multi-attribute motion data can be generated in many applications/ devices, such as motion capture devices and animations. It can have dozens of attributes, thousands of rows, and even similar motions can have different durations and different speeds at corresponding parts. There are no row-to-row correspondences between data matrices of two motions. To be classified and recognized, multi-attribute motion data of different lengths are reduced to feature vectors by using the properties of singular value decomposition (SVD) of motion data. The reduced feature vectors of similar motions are close to each other, while reduced feature vectors are different from each other if their motions are different. By applying support vector machines (SVM) to the feature vectors, we efficiently classify and recognize real-world multi-attribute motion data. With our data set of more than 300 motions with different lengths and variations, SVM outperforms classification by related similarity measures, in terms of accuracy and CPU time. The performance of our approach shows its feasibility of real-time applications to real-world data.

Segmentation and recognition of motion streams by similarity search

Fast and accurate recognition of motion data streams from gesture sensing and motion capture devices has many applications and is the focus of this article. Based on the analysis of the geometric structures revealed by singular value decompositions (SVD) of motion data, a similarity measure is proposed for simultaneously segmenting and recognizing motion streams. A direction identification approach is explored to further differentiate motions with similar data geometric structures. Experiments show that the proposed similarity measure can segment and recognize motion streams of variable lengths with high accuracy, without knowing beforehand the number of motions in a stream.

Segmentation and recognition of motion capture data stream by classification

Three dimensional human motions recorded by motion capture and hand gestures recorded by using data gloves generate variable-length data streams. These data streams usually have dozens of attributes, and have different variations for similar motions. To segment and recognize motion streams, a classification-based approach is proposed in this paper. Classification feature vectors are extracted by utilizing singular value decompositions (SVD) of motion data. The extracted feature vectors capture the dominating geometric structures of motion data as revealed by SVD. Multi-class support vector machine (SVM) classifiers with class probability estimates are explored for classifying the feature vectors in order to segment and recognize motion streams. Experiments show that the proposed approach can find patterns in motion data streams with high accuracy.

A similarity measure for motion stream segmentation and recognition

Recognition of motion streams such as data streams generated by different sign languages or various captured human body motions requires a high performance similarity measure. The motion streams have multiple attributes, and motion patterns in the streams can have different lengths from those of isolated motion patterns and different attributes can have different temporal shifts and variations. To address these issues, this paper proposes a similarity measure based on singular value decomposition (SVD) of motion matrices. Eigenvector differences weighed by the corresponding eigenvalues are considered for the proposed similarity measure. Experiments with general hand gestures and human motion streams show that the proposed similarity measure gives good performance for recognizing motion patterns in the motion streams in real time.

Segmentation and recognition of multi-attribute motion sequences

In this work, we focus on fast and efficient recognition of motions in multi-attribute continuous motion sequences. 3D motion capture data, animation motion data, and sensor data from gesture sensing devices are examples of multi-attribute continuous motion sequences. These sequences have multiple attributes rather than only one attribute as time series data has. Motions can have different rates and durations, and the resulting data can thus have different lengths. Also, motion data can have noises due to transitions between successive motions. Hence, traditional distance measuring approaches used for time series data (such as Euclidean distances or dynamic time-warped distances) are not suitable for recognition in multi-attribute motion sequences. Hence, we have defined a similarity measure based on the analysis of singular value decomposition (SVD) properties of similar multi-attribute motions. A five-phase algorithm has then been proposed that gives good pruning power by exploiting the proximity of continuous motion data. We experimented this algorithm with data from different sources: 3D motion capture devices, animation motions, and CyberGlove gesture sensing device. These experiments show that our algorithm can segment and recognize long motion streams with high accuracy and in real time without knowing beforehand the number of motions in a stream.

Indexing of variable length multi-attribute motion data

Haptic data such as 3D motion capture data and sign language animation data are new forms of multimedia data. The motion data is multi-attribute, and indexing of multi-attribute data is important for quickly pruning the majority of irrelevant motions in order to have real-time animation applications. Indexing of multi-attribute data has been attempted for data of a few attributes by using R-tree or its variants after dimensionality reduction. In this paper, we exploit the singular value decomposition (SVD) properties of multi-attribute motion data matrices to obtain one representative vector for each of the motion data matrices of dozens or hundreds of attributes. Based on this representative vector, we propose a simple and efficient interval-tree based index structure for indexing motion data with large amount of attributes. At each tree level, only one component of the query vector needs to be checked during searching, comparing to all the components of the query vector that should get involved if an R-tree or its variants are used for indexing. Searching time is independent of the number of pattern motions indexed by the tree, making the index structure well scalable to large data repositories. Experiments show that up to 91∼93% irrelevant motions can be pruned for a query with no false dismissals, and the query searching time is less than 30 μ <i>s</i> with the existence of motion variations.

Hierarchical indexing structure for 3d human motions

Content-based retrieval of 3D human motion capture data has significant impact in different fields such as physical medicine, rehabilitation, and animation. This paper develops an efficient indexing approach for 3D motion capture data, supporting queries involving both sub-body motions (e.g., Find similar knee motions) as well as whole-body motions. The proposed indexing structure is based on the hierarchical structure of the human body segments consisting of independent index trees corresponding to each sub-part of the body. Each level of every index tree is associated with the weighted feature vectors of a body segment and supports queries on sub-body motions and also on whole-body motions. Experiments show that up to 97% irrelevant motions can be pruned for any kind of motion query while retrieving all similar motions, and one traversal of the index structure through all index trees takes on an average 15 μsec with the existence of motion variations.

Feature Selection for Classification of Variable Length Multiattribute Motions

As a relatively new type of multimedia, captured motion has its specific properties. The data of motions has multiple attributes to capture movements of multiple joints of a subject, and has different lengths for even similar motions. There are no row-to-row correspondences between data matrices of two motions. To be classified and recognized, multi-attribute motion data of different lengths are reduced to feature vectors by using the properties of Singular Value Decomposition (SVD) of motion data in this Chapter. Different feature selection approaches are explored, and by applying Support Vector Machines (SVM) to the feature vectors, we can efficiently classify and recognize real world multi-attribute motion data. With our datasets of hundreds of 3D motions with different lengths and variations, classification by SVM is compared with classification by related similarity measures, in terms of accuracy and CPU time.

Indexing of Motion Capture Data for Efficient and Fast Similarity Search

As motion capture systems are increasingly used for motion tracking and capture, and more and more surveillance cameras are installed for security protection, more and more motion data, including 3D motion data becomes available, making it important to index motion data for quick retrieval of similar motions. This paper proposes a simple and efficient tree structure for indexing motion data with dozens of attributes. Feature vectors are extracted for indexing by using singular value decomposition (SVD) properties of motion data matrices. By having similar motions with large variations indexed together, searching for similar motions of a query needs only one node traversal at each tree level, and only one feature needs to be considered at one tree level. Experiments with real hand gestures, arm motions and full body motions show that the majority of irrelevant motions can be pruned while retrieving all similar motions, and the traversal of an indexing tree for a querytakes only a few microseconds.

Motion Stream Segmentation and Recognition by Classification

This paper proposes a classification-based approach to segmenting and recognizing patterns in motion signals. Feature vectors are extracted based on singular value decomposition (SVD) for classification. Multi-class support vector machine (SVM) classifiers with class probability estimates are explored for segmenting and recognizing motion streams. Experiments show that the proposed approach can find patterns in the multi-attribute motion streams with high accuracy