Thesis or Dissertation コミュニケーション支援を目的とした非言語動作 の検出に関する基礎検討

小原, 崇司

In this paper, we propose a nonverbal motion analyze for communication support. Two volunteer subjects during their conversation are tracked by the Kinect v2 sensors and got their 3-D skeleton information. We was performed noise cutting from getting 3-D information andprocessing principal component analysis (PCA), wavelet multi-resolution analysis to decompose each time sequence about total of 65 people (110 movie ,length 300s). We build the average motion model (AMM) from processing dataset, and attempted motion analysis which estimated as nonverbal. In addition, we asked 25 people to evaluate 10 videos and got training data to detect. As a result , the proposed system attained precision=52.2% and recall=67.6%Key Words:Kinect v2 ,Motion analyze, MRA

Number of accesses :  

Other information