Research On The Capture Data Retrieval Of Human Motion Graph Based On Hand Drawing

Due to the rapid development of internet of things technology in recent years,intelligent hardware devices on the market now appear thick and fast,people enjoy the convenience and benefits in the information age,at the same time,various kinds of massive data have been formed.People use kinds of technical method to obtain information and to deal with information data.Motion capture technology has a good application prospect in the field of movie,animation,game and virtual reality.The application value of motion capture device is becoming more and more prominent,and it drives the development of large-scale human and also renew the object motion database.But,with all kinds of human motion data growing in recent years,it is difficult to retrieve animated movements which can meet specific requirements.At the same time,the reuse of databases built with expensive device capture and a large amount of updated motion capture data has become a hot research topic.Motion retrieval technology in recent years has also become the main focus in the field of motion capture animation.Therefore,using freehand-sketching human motion series to retrieve captured data is a more convenient and effective method for human motion retrieval.In this paper,a motion retrieval method based on the combination of automatic encoder and dynamic time warping is proposed.It mainly contains two parts of the work:(1)A method to extract the feature of motion sequence by using the representative frame and the automatic encoder is proposed.In the stage of motion sequence feature learning,according to the motion sequence in the motion database,the first step is to use the fuzzy clustering to obtain the representative frame of motion,and then to establish the key frame image set.Using the neural network to learn the key frame image set,the automatic feature coder is obtained.Automatic encoder to extract the feature of each key frame image,then establish the motion feature database.(2)A retrieval framework using hand drawing query and DTW motion sequence matching is proposed.In the motion retrieval phase,according to the motion sequences to be queried in the motion database,use the automatic encoder obtained in stage(1)to extract the feature of each key frame image,and then calculating the similarity between the motion to be queried and the database in the motion based on the dynamic time warping method,and sort the retrieval results according to the similarity measure.The above method is used in this paper,and part of the human sketched motion sequences is used to carry out the matching retrieval experiments with human motion data in the human motion capture database.The experiment validates the effectiveness of the proposed method.