研究目的
To develop a new healthcare monitoring system for vulnerable people that uses multispectral data processing to overcome environmental challenges and improve activity recognition.
研究成果
The developed healthcare monitoring system effectively uses multispectral data and chromatic methodology to improve activity recognition for vulnerable people. The novel human silhouette identification method and COGD descriptor show robustness against disturbances, with better performance compared to traditional methods. Future work could focus on real-world deployment and further optimization.
研究不足
The system was tested in a controlled laboratory environment, which may not fully represent real-world noisy and dynamic conditions. The PIR sensors have a recovery period that can affect tracking, and the method requires initial calibration of spatial regions.
1：Experimental Design and Method Selection:
The system uses chromatic methodology to process multispectral data from visible and infrared sensors. It involves a novel human silhouette identification method and a new chromatic descriptor (COGD) for activity recognition.
2：Sample Selection and Data Sources:
Data sets were generated by four participants performing ten daily living activities in a controlled laboratory environment. Over 175,000 foreground images were created.
3：List of Experimental Equipment and Materials:
Sensing head with CMOS Sony image sensor, fisheye lens, three Panasonic PIR sensors, Broadcom microcontroller (64bit ARMv8 quad core Cortex A53 processor at
4：2GHz), Microsoft Kinect sensor for ground truth, and a PC for data processing. Experimental Procedures and Operational Workflow:
The sensing head was installed overhead to capture images and PIR signals. Data was processed using ViBe foreground detector, human silhouette identification with probability models, and COGD descriptor for feature extraction. Performance was evaluated using k-nearest neighbor classifier.
5：Data Analysis Methods:
Percentage of correct classification (PCC) and confusion matrices were used to assess performance under different disturbance levels.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
