Research On High Performance Retrieval Technology Of Motion Capture Data

Motion capture is a method of digitally characterizing the temporal and spatial structure of human motion.As the amount of motion capture data increases,it is becoming more and more important to retrieve data in order to improve the performance of information access.The retrieval technique provides a way to search for similar motions in a motion capture database,identifying that the query action is one of several motion classes.This paper implements and evaluates five traditional motion capture data retrieval techniques based on PCA,feature function and DTW.A SK-means-based motion capture data retrieval method that combines accuracy and efficiency is proposed.main tasks as follows:1.The five motion capture data retrieval techniques based on PCA,feature function and DTW are implemented and evaluated.It is found that the number of PCA-based missed reports is the least but there are a large number of false positives(close to 90%),which indicates that the two technologies are incorrectly The unmatched categories were retrieved.Feature-based and DTW-based technologies perform better than PCA-based technologies.Among them,the motion capture data retrieval technology based on feature function has higher efficiency,but it is higher than the DTW-based retrieval technology for false negatives and false positives.A motion capture retrieval algorithm that combines efficiency and accuracy is proposed.The method organizes motion capture data based on statistical k-means(SK-means),democratic decision-making,unsupervised learning,visual key frame extraction,etc.,and realizes intuitive retrieval by browsing thumbnails of semantic key frames.In order to efficiently retrieve motion capture data,we have taken three steps.First,the new SK-means algorithm is used to cluster the motion capture data to obtain the basic categories.Then perform character matching.The system learns the user's retrieval information during the retrieval process,updates the retrieval success rate of each data,and sorts the search results according to the retrieval success rate through democratic decision-making,thereby improving the accuracy of the search results.Finally,a semantic generalized thumbnail is generated.This method uses a key frame extraction algorithm based on visual data analysis.Experiments show that this method can effectively organize and retrieve massive motion capture data.