紀要論文 動的特徴を用いた歩容認証:RNN 及び SVM の性能比較

熊埜御堂, 裕太

内容記述
This paper presents a gait recognition system that uses recurrent neural networks (RNNs)and support vector machines (SVMs) for identifying individuals. Our system extracts thespatiotemporal features of distances between the waist and various joint positions obtainedby a Kinect sensor. These spatiotemporal features are invariant for a walking subject. Toverify our system performance, we conducted tests using the data of 12 individuals. Thedata were divided into two datasets for training and testing. The RNNs and SVMs weretrained for classification using the training dataset. SVMs achieved an average accuracy ofover 99% for the test dataset, whereas the average accuracy of RNNs was 94%.Kew Word : RNN,SVM,Kinect,gait recognition
本文を読む

http://repo.lib.hosei.ac.jp/bitstream/10114/13571/1/15R4115.pdf

このアイテムのアクセス数:  回

その他の情報