Gesture Analysis in Conversational Interaction Using a Time-Series Data Mining Approach(<Special Issue>A Quantitative Empirical Study on Interpersonal Communication)

DOI Web Site

Bibliographic Information

Other Title
  • 時系列データマイニングを援用した会話インタラクションにおけるジェスチャ分析の支援(<特集>対人コミュニケーションに関する定量的実証研究)
  • 時系列データマイニングを援用した会話インタラクションにおけるジェスチャ分析の支援
  • ジケイレツ データマイニング オ エンヨウ シタ カイワ インタラクション ニ オケル ジェスチャ ブンセキ ノ シエン

Search this article

Abstract

This paper introduced examples of gesture analysis by using a time-series data mining system. A time-series data mining system is able to reduce time and effort to annotate a large amount of nonverbal behavior data. At first, we obtain the trajectory of hand motion as multi-dimensional time-series data by using motion capture system. Second, each frame samples in the time-series data is classified to no motion patterns (such case that hand is set on home position) and motion patterns including gestures automatically. Third, frequent gesture patterns are extracted in time series by using motif (pattern) discovery algorithm. We applied these data mining methods to analyze gestures which are used in an explanation task. In the task, two participants explain the contents of animation to one participant. We collected datasets of 8 groups which are composed of 24 participants. Results of data mining show that amount of gestures in the task and timing to use gestures are different among groups and we can classify the timing in 8 groups to 3 types. From results of extracting of frequent gesture patterns, we show that 3 kind of gestures are used commonly in some groups.

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top