Scholarlink

A significant advancement in sensor-based technologies has led to a rapid expansion of applications for the Internet of Things (IoT), which can be used to build any real-time monitoring system. This expansion can be attributed to the significant advancement in sensor-based technologies. Nowadays, an increasing number of elderly people live alone in various locations worldwide, and it is essential to monitor the current state in which their health function or activity is critical. Human activity recognition (HAR) technology based on wearable sensors has many applications, including healthcare, fitness, smart homes, and surveillance. However, despite extensive computational research on HAR, multi-sensor-based activity recognition has numerous open challenges. One of these challenges is recognising complex temporal data, identifying discriminate feature vectors from multi-modal data and reducing data dimensionality. These issues necessitate extensive research. The Internet of Things (IoT) is the foundation for this paper’s proposed human activity monitoring model. Using innovative sensor-based technologies, the model is designed to examine the actions of elderly people continuously. This model collects vital information using intelligent sensors or devices powered by the Internet of Things (IoT). Machine learning algorithms analyse this data to identify potential human behaviour risks that can be deduced from their actions. The proposed work’s utility is assessed using a dataset made public by UCI. The findings of the experiments, which made use of several different metrics in addition to a parallel coordinate plot (PCP), indicate that the proposed ML-based HAR framework outperforms several state-of-the-art ML techniques and has the potential to achieve a maximum classification accuracy of 96.30%. Following the evaluation of the proposed model, the SVM was found to have committed the best accuracy of 96.30%, which is quite helpful for our objectives. In addition, SVM performed significantly better than every other machine learning algorithm tested for this investigation.

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

A Competent Human Activity Detection Using Integrated Wearable Sensor Data and Machine Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Veeramuthu Venkatesh,
  • Pethuru Raj,
  • Roshitha Nedium,
  • Jahnavi Edara,
  • Kalluru Amarnath Reddy,
  • R. Anushiadevi

摘要

A significant advancement in sensor-based technologies has led to a rapid expansion of applications for the Internet of Things (IoT), which can be used to build any real-time monitoring system. This expansion can be attributed to the significant advancement in sensor-based technologies. Nowadays, an increasing number of elderly people live alone in various locations worldwide, and it is essential to monitor the current state in which their health function or activity is critical. Human activity recognition (HAR) technology based on wearable sensors has many applications, including healthcare, fitness, smart homes, and surveillance. However, despite extensive computational research on HAR, multi-sensor-based activity recognition has numerous open challenges. One of these challenges is recognising complex temporal data, identifying discriminate feature vectors from multi-modal data and reducing data dimensionality. These issues necessitate extensive research. The Internet of Things (IoT) is the foundation for this paper’s proposed human activity monitoring model. Using innovative sensor-based technologies, the model is designed to examine the actions of elderly people continuously. This model collects vital information using intelligent sensors or devices powered by the Internet of Things (IoT). Machine learning algorithms analyse this data to identify potential human behaviour risks that can be deduced from their actions. The proposed work’s utility is assessed using a dataset made public by UCI. The findings of the experiments, which made use of several different metrics in addition to a parallel coordinate plot (PCP), indicate that the proposed ML-based HAR framework outperforms several state-of-the-art ML techniques and has the potential to achieve a maximum classification accuracy of 96.30%. Following the evaluation of the proposed model, the SVM was found to have committed the best accuracy of 96.30%, which is quite helpful for our objectives. In addition, SVM performed significantly better than every other machine learning algorithm tested for this investigation.