Automatic Extraction of Semantic Action Features

Thanh, Tran Thang; Chen, Fan; Kotani, Kazuhiko; Le, Bac

doi:10.1109/sitis.2013.35

Cited by 4 publications

(3 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The surveyed papers exploit different kinds of motion features that are more or less suitable for the specific application purpose, such as retrieval of similar motions, generation of new realistic motions, person identification according to the style of movement, and action‐based annotation of an unlabeled motion. In particular, the extracted motion features of all the papers can be summarized into the following groups: joint angles (rotations) , distances between joints , relative joint velocity and acceleration , raw 3D joint coordinates , and a combination of the aforementioned . To evaluate discriminative power of various features and comparison functions for different kinds of applications, we select the five representative approaches, which are summarized in Table .…”

Section: Case Studiesmentioning

confidence: 99%

“…To describe characteristic aspects of motion data, various kinds of motion features can be extracted from the raw 3D joint coordinates, such as joint angles (rotations) , distances between joints , relative joint velocity and acceleration , and their combinations . The extracted features provide a necessary abstraction of motion data by emphasizing important motion components while suppressing everything insignificant to our interest.…”

Section: Introductionmentioning

confidence: 99%

“…As soon as appropriate features are chosen and ex‐ tracted, it is essential to define the way of their comparison. The extracted features can be compared on the basis of nearest‐neighbor search (e.g., provided by index structures) , term frequency – inverse document frequency (TF‐IDF) , machine learning (e.g., provided by support vector machines (SVMs) or artificial neural networks ), and probabilistic methods (e.g., provided by Markov models or Bayesian networks ). The TF‐IDF and nearest‐neighbor approaches return a list of the most similar motions with respect to a query motion example as the comparison output, while machine learning and probabilistic methods return a set of classes to which the query example is classified.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Assessing similarity models for human‐motion retrieval applications

Valčík

Sedmidubský

Zezula

2015

Computer Animation & Virtual

View full text Add to dashboard Cite

The development of motion capturing devices poses new challenges in the exploitation of human-motion data for various application fields, such as computer animation, visual surveillance, sports, or physical medicine. Recently, a number of approaches dealing with motion data have been proposed, suggesting characteristic motion features to be extracted and compared on the basis of similarity. Unfortunately, almost each approach defines its own set of motion features and comparison methods; thus, it is hard to fairly decide which similarity model is the most suitable for a given kind of human-motion retrieval application. To cope with this problem, we propose the human motion model evaluator, which is a generic framework for assessing candidate similarity models with respect to the purpose of the target application. The application purpose is specified by a user in form of a representative sample of categorized motion data. Respecting such categorization, the similarity models are assessed from the effectiveness and efficiency points of view using a set of space-complexity, information-retrieval, and performance measures. The usability of the framework is demonstrated by case studies of three practical examples of retrieval applications focusing on recognition of actions, detection of similar events, and identification of subjects.

show abstract