Scholarlink

<p>In healthcare, the term activity of daily living (ADL) refers to the basic activities that humans perform daily to maintain independence and autonomy. A daily routine is an ordered set of ADLs that reflects a person’s lifestyle and wellness. This article proposes a weighted spatial-temporal features method (WSTF) based on fusion adaptive resonance theory (ART) to discover daily routines. WSTF integrates multiple activity attributes, including starting time, end time, and location, to generate spatial-temporal patterns of ADL. A weight assignment scheme is proposed to determine the weight of activity patterns based on the probability of activity transition on the day to distinguish the importance of the activity. Then daily routines are learned from a set of weighted spatial-temporal ADL sequences. Experiments are conducted on data sets collected by two smart home projects, learning 3 and 14 routines from the 20-day Orange4Home and 220-day Aruba continuous activity streams, respectively. The routines we obtain are more consistent, with fewer clusters and fewer errors than the baseline methods. The results of this study provide a basis for the quantification of daily routines and have great potential for the development of real-world applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Daily routines discovery based on weighted spatial-temporal features

  • 期刊论文

  • 发表日期：

    2025年2月3日

查看全文
  • Xinjing Song,
  • Yanjiang Wang

摘要

In healthcare, the term activity of daily living (ADL) refers to the basic activities that humans perform daily to maintain independence and autonomy. A daily routine is an ordered set of ADLs that reflects a person’s lifestyle and wellness. This article proposes a weighted spatial-temporal features method (WSTF) based on fusion adaptive resonance theory (ART) to discover daily routines. WSTF integrates multiple activity attributes, including starting time, end time, and location, to generate spatial-temporal patterns of ADL. A weight assignment scheme is proposed to determine the weight of activity patterns based on the probability of activity transition on the day to distinguish the importance of the activity. Then daily routines are learned from a set of weighted spatial-temporal ADL sequences. Experiments are conducted on data sets collected by two smart home projects, learning 3 and 14 routines from the 20-day Orange4Home and 220-day Aruba continuous activity streams, respectively. The routines we obtain are more consistent, with fewer clusters and fewer errors than the baseline methods. The results of this study provide a basis for the quantification of daily routines and have great potential for the development of real-world applications.