{
  "paper_id": "2016664477f070bb9a6cfdbf9131339c",
  "paper_title": "Representative Body Points on Top-View Depth Sequences for Daily Activity Recognition",
  "paper_authors": [
    "Shu-Chun Lin",
    "An-Sheng Liu",
    "Tang-Wei Hsu",
    "Li-Chen Fu"
  ],
  "publication_date": "",
  "publication_venue": "2015 IEEE International Conference on Systems, Man, and Cybernetics",
  "abstract": "In this paper, a novel feature for activity recognition from vertical top-view depth image sequences is firstly proposed. Most of previous works are focusing mainly on the side-view depth or color image sequences, which unfortunately may encounter occlusion problems. Therefore, top-view camera setting is adopted in our research. Based on the idea of computed tomography (CT) from medical imaging, the depth images are segmented to different layer along the transverse plane. The representative body points which are found from the centroids of the regions on each slice. And those points will be a meaningful descriptor for the activity posture. Dynamic time warping algorithm is also applied to address the different sequence length problem. Finally, a SVM classifier is trained to classify our activities. To verify our performance, a new Top-View 3D Daily Activity Dataset is constructed. In our experiments, a challenging cross-subject test is conducted, and the performance of our representative body points is demonstrated. The result shows that the accuracy can achieve up to 97%, which is promising while being compared with those from the state-of-the-art methods in the literature.",
  "keywords": [
    "action recognition",
    "top-view",
    "depth",
    "dynamic time warping"
  ],
  "paper_institutions": [
    "Department of Electrical Engineering, National Taiwan University"
  ],
  "output_language": "zh",
  "provider": "doubao",
  "model": "ark-code-latest",
  "prompt_template": "simple",
  "extracted_at": "2026-08-16T19:46:54.703401Z",
  "template_tag": "simple",
  "summary": "### 摘要\n本文针对俯视深度图像序列中的日常活动识别问题，提出了一种基于代表性身体点的新特征。相较于传统的侧视深度或彩色图像序列容易发生遮挡，本文采用俯视摄像头设置，借鉴医学成像中的计算机断层扫描（CT）思想，将深度图像沿横断面分层，每层区域质心作为代表性身体点，构成姿态描述符。同时，采用动态时间规整（DTW）算法解决序列长度不一致的问题，并训练支持向量机（SVM）分类器进行活动识别。为验证方法，构建了新的俯视3D日常活动数据集，包含6种活动，由9名受试者执行，并进行挑战性的跨受试者测试。实验结果表明，所提方法准确率达97.7%，优于现有最优方法STOP（92.6%），尤其在区分sit down和squat等相似活动时表现更佳，证明该特征能有效缓解自我遮挡和模糊序列问题，适用于监控和老年人健康监护等真实场景。\n\n### 关键要点\n- 论文提出了一种新的活动识别特征，即代表性身体点，用于俯视深度图像序列。\n- 该方法基于医学CT成像思想，将深度图像分层处理，并为每层区域计算质心作为代表性点。\n- 采用动态时间规整（DTW）算法对不同长度的活动序列进行对齐，为后续分类提供统一特征向量。\n- 使用SVM分类器对提取的特征序列进行分类识别。\n- 构建了新的Top-View 3D Daily Activity Dataset，包含行走、站起、坐下、深蹲、打麻将和跌倒等6种日常活动。\n- 数据集由9名受试者（7男2女）执行，每项活动重复3次，共162个序列，身高范围1.6-1.9米，增加了类内差异。\n- 实验采用跨受试者方案，训练集5人、测试集4人，重复3次取平均结果。\n- 代表性身体点方法平均准确率为97.7%，优于STOP方法的92.6%。\n- 混淆矩阵显示，代表性身体点方法在sit down和squat上的区分准确率显著高于STOP，混淆更少。\n- 深度值范围500-3700 mm，每幅图分为3层（头层、头层+1、头层+2），层间距220 mm，仅考虑头层和其下方两层，每层分别取1、2、2个代表点。\n- 该方法的优势在于能处理自我遮挡和噪声，且适用于监控场景，无需用户面向摄像头，保护隐私。\n- 局限性包括依赖深度传感器精度，对严重噪声仍需处理，且目前仅在自建数据集上验证，未与其他更大规模数据集比较。",
  "available_templates": [
    "deep_read",
    "simple"
  ]
}