基於LSTM模型的動作識別及KNN分類的地理位置識別之研究

Research on the LSTM Model Based Action Recognition and the KNN Classification Based Geolocation Recognition

黃懷逸(HUAI-YI HUANG)

指導教授 ： 陳忠信

朝陽科技大學/資訊學院/資訊與通訊系/碩士(2020年)

摘要

本文包含兩研究主題，第一個主題是基於長短期記憶網路(Long Short-Term Memory, LSTM)，有效地將連續的人體關鍵點分類成動作。在動作識別實驗的樣本以柔道的3個競技動作搭配6個一般動作來做動作識別，實際結果可得92%的準確率。第二個主題是基於K-近鄰演算法(K-Nearest Neighbor Classification, KNN)，KNN是人工智慧中的一種分類方法，透過特徵值比較快速的預測一個樣本的可能分類。在位置識別實驗的樣本以台北市為區域，以知名社群網路的地理位置點(geographical points)。實驗的結果顯示我們提出權重KNN方法準確率可達99.5%以上。

關鍵字

OpenPose ； 遞歸神經網路 ； 長短期記憶 ； 動作識別 ； 運動識別 ； 柔道 ； K-近鄰演算法 ； 社交網路服務 ； 地理位置預測

並列摘要

This paper contains two research topics. The first topic is to effectively classification continuous human body key points into actions based on Long Short-Term Memory (LSTM). In the action recognition experiment, the three competitive actions of Judo combined with six general actions are recognized. The actual result can achieve 92% accuracy. The second topic is based on K-Nearest Neighbor Classification (KNN). KNN is a classification method in artificial intelligence, which can quickly predict the possible classification of a sample through feature values. The sample in the location recognition experiment uses Taipei City as the area and uses geographic points of well-known social networks. The experimental results show that the accuracy of our proposed weighted KNN method can reach more than 99.5%.

並列關鍵字

OpenPose ； RNN ； LSTM ； Action Recognition ； Motion Recognition ； Judo ； KNN ； Social Network Services ； Geographic Location Prediction

參考文獻

[1] P. Zhang, C. Lan, J. Xing, W. Zeng, J. Xue, and N. Zheng, "View Adaptive Recurrent Neural Networks for High Performance Human Action Recognition from Skeleton Data, "IEEE International Conference on Computer Vision, 2017, pp. 2117-2126.

[2] H. Zhang, Y. Song and Y. Zhang, "Graph Convolutional LSTM Model for Skeleton-Based Action Recognition," IEEE International Conference on Multimedia and Expo (ICME), Shanghai, China, 2019, pp. 412-417

[3] S. Song, C. Lan, J. Xing, W. Zeng and J. Liu, "Spatio-Temporal Attention-based LSTM Networks for 3D Action Recognition and Detection," in IEEE Transactions on Image Processing, Vol. 27, No. 7, pp. 3459-3471.

[4] S. Jun and Y. Choe, "Deep Batch-Normalized LSTM Networks with Auxiliary Classifier for Skeleton based Action Recognition,"IEEE International Conference on Image Processing, Applications and Systems (IPAS), Sophia Antipolis, France, 2018, pp. 279-284.

[5] X. Ouyang et al., "A 3D-CNN and LSTM Based Multi-Task Learning Architecture for Action Recognition," IEEE Access, Vol. 7, pp. 40757-40770.