During the process of multi task human motion behavior feature recognition, it is susceptible to the influence of small samples, resulting in catastrophic forgetting and poor ability to handle variability, leading to a decrease in feature recognition performance. Therefore, this article proposes a multi task behavior feature recognition method based on chaotic invariants. This method uses the C-C method to estimate time delay, reconstruct the phase space of human motion behavior, and convert one-dimensional time series data into high-dimensional phase space; Extracting chaotic invariants representing human behavior from the reconstructed phase space as features, compressing high-dimensional raw motion data into low dimensional feature vectors, preserving key information and reducing computational complexity, utilizing the powerful feature expression ability and invariance of chaotic invariants to provide consistent feature representation between different tasks, thereby supporting multi task recognition of human motion behavior and effectively improving variability processing capabilities. By using the multi task learning method of support vector machine to classify and recognize the extracted chaotic invariant features, accurate recognition of multi task human motion behavior features can be achieved. The experimental results show that this method can efficiently recognize human motion features, accurately capture subtle changes in joints, and perform well in both single task and complex multi task, cross individual scenarios. Extracting chaotic invariants can effectively address the complexity and diversity of motion data, improving recognition accuracy and stability.
QiaoYi, QuYi. Human behavior recognition method based on adaptive fusion weights[J]. Computer Engineering and Design, 2023, 44(3): 845-851.
[9]
NguyenT, PhamD T, VuH,et al.A robust and efficient method for skeleton-based human action recognition and its application for cross-dataset evaluation[J]. IET Computer Vision, 2022, 16(8): 709-726.
HuangYu-chen, LuoJian, YangQiang. Research on gait recognition based on spatiotemporal image fusion and multi task classification network[J]. Optical Technology, 2023, 49(1): 97-104.
[12]
RahulK, ShailenderK.Survey on artificial intelligence-based human action recognition in video sequences[J]. Optical Engineering, 2023, 62(2): 023102-1-21.
LiuShu-yong, ChaiKai, WeiYun-peng, et al. Chaos signal recognition method based on phase space reconstruction and convolutional neural network[J]. Journal of Naval Engineering University, 2023, 35(3): 59-68.
ZhangLi-guo, LiuWan, ZhangShu-qing, et al. Chaos singular spectrum analysis and application based on phase space reconstruction improved algorithm[J]. Journal of Metrology, 2021, 42(10): 1299-1306.
YuanZe-shi, QiQing-qing, ZhuXiao-hua. Method for selecting attractors of chaotic systems[J]. Journal of Nanjing University of Science and Technology, 2023, 47(6): 836-841.
ZhengZhen, ZouJun-ying. Research on motion data classification based on chaotic correlation dimension and decision tree[J]. Computer Simulation, 2022, 39(10): 327-330, 424.
ShaoHui-ting, YangQi-gui. Complex dynamics study of six dimensional hyperchaotic systems with four positive Lyapunov exponents[J]. Journal of Guangxi Normal University (Natural Science Edition), 2022, 40(5): 433-444.
GaoYun-bao, SunYu-quan. Fuzzy C-means clustering small data method for calculating Lyapunov exponent [J]. Journal of Northeast Normal University (Natural Science Edition), 2021, 53(1): 38-42.
TaoJia-gui, HuSheng-nan, DaiJian-zhuo, et al. Research on human behavior recognition method based on improved machine learning[J]. Electric Transmission, 2023, 53(11): 69-76.