研究目的
To distinguish and detect vital signs of users in a domestic environment using machine learning techniques supplementing UWB radar, building on previous work for intelligent location awareness.
研究成果
The proposed method achieves high accuracy (98.3%) in distinguishing vital signs and room locations using UWB radar and machine learning, with potential applications in monitoring activity levels for elderly care. Future work will extend to multiple occupants and longer time intervals.
研究不足
The experiment is conducted in a specific domestic environment with a single UWB device and limited to one person; it may not generalize to multiple occupants or different settings. Overfitting occurs with high-dimensional data (e.g., 30% training data), and the system relies on prior training for localization.
1：Experimental Design and Method Selection:
The experiment uses UWB radar to detect vital signs (breathing and heartbeat) in different rooms of a house, with signal analysis via STFT for time-frequency patterns and classification using MC-SVM.
2：Sample Selection and Data Sources:
Data is collected from a semi-detached house in Essex, UK, with four rooms (living room, kitchen, dining room, bathroom), involving persons present or absent, with static environment.
3：List of Experimental Equipment and Materials:
UWB radar module (Time Domain PulsON 410), Raspberry Pi for data storage, Matlab R2016b for processing, Intel Core i7 processor, Windows 7 OS, NVIDIA GPU.
4：Experimental Procedures and Operational Workflow:
UWB signals are transmitted and received, pre-processed to remove direct path interference, analyzed with STFT using a Hamming window, and features are classified with MC-SVM trained on time stamp and localization data.
5：Data Analysis Methods:
Statistical analysis of classification results (correct rate, error rate, sensitivity, specificity, PPV, NPV) using MC-SVM, with thresholds for vital sign distinction based on known rates.
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
