A Geometric Approach for Cross-View Human Action Recognition using Deep Learning

• Published in: 2019 11th International Symposium on Image and Signal Processing and Analysis (ISPA) pp. 258 - 263
• Main Authors: Papadakis, Antonios, Mathe, Eirini, Spyrou, Evaggelos, Mylonas, Phivos
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2019
• Subjects: Cameras; Convolutional Neural Networks; Feature extraction; Human Activity Recognition; Skeleton; Three-dimensional displays; Training; Transforms; Two dimensional displays
• ISSN: 1849-2266
• DOI: 10.1109/ISPA.2019.8868717

In this paper we present an approach for the recognition of human actions which is based on a deep Convolutional Neural Network architecture. More specifically, 3D skeletal joint information is used to create 2D (image) representations. To compensate for potential viewpoint changes, these images are pre-processed using geometric transformations. Then, they are transformed to the spectral domain using well-known transforms. We focus on actions that are close to activities of daily living (ADLs), yet we evaluate our approach using a large-scale action dataset. We cover single-view, cross-view and cross subject cases and thoroughly discuss experimental results and the potential of our approach.