查詢結果分析
來源資料
頁籤選單縮合
題 名 | An Efficient Temporal Model for Action Recognition Using Multivariate Linear Prediction=使用多變數線性預測之人類動作辨識的時間資訊模型 |
---|---|
作 者 | 林晉安; 林彥宇; 鄭士康; | 書刊名 | 中華民國資訊學會通訊 |
卷 期 | 16:1 2013.03[民102.03] |
頁 次 | 頁1-19 |
分類號 | 312.13 |
關鍵詞 | 多變數線性預測; 人類動作辨識; 時間資訊模型; |
語 文 | 英文(English) |
英文摘要 | To recognize temporally extended actions, it is profitable to introduce high-order temporal dependence into recognition tasks. However, such thing is hardly achieved, when the conventional models such as Hidden Markov Model and Conditional Random Feild are employed. In this paper, we propose multivariate linear predication to effciently exploit the dependence. In addition, our method take advantages of bag-of-word representations, which may contain considerable noise but has shown excellent results in previous work. To show its applicability, we experiment not only on video datasets including KTH and UCF but on skeleton datasets such as MSR 3D action and UCF Kinect. In most of them, our method gets superior performance than the state-of-the-art methods. |
本系統中英文摘要資訊取自各篇刊載內容。