Self-collected IR-UWB Dataset for Human Activity Recognition

This is a IR-UWB radar dataset for human activity recognition.  The Xethru X4 UWB radar is used for data collection. Five volunteers, including three men and two women are invited to perform six different activities relevant to daily life, including Turn around, Pacing, Jump up, Wave hands, Fall down and Sit & stand. To enhance the variability of the dataset, volunteers were encouraged to perform activities based on their personal routines without adhering to standardized restrictions. Initially, the raw data undergoes a process of noise reduction and clutter suppression. This involves distance selection and elimination of direct current (DC) components. Subsequently, TR feature maps are extracted by segmenting the data using partially overlapping sliding time windows. We also provide a *.csv file for 5-fold cross-validation experiments based on 5 participants.本数据集为用于人体活动识别的脉冲超宽带（IR-UWB）雷达数据集。数据采集采用Xethru X4超宽带（UWB）雷达完成。本次实验邀请了5名志愿者（3男2女）完成6种日常相关的典型活动，分别为转身、踱步、起跳、挥手、跌倒以及坐下与起立。为提升数据集的多样性，志愿者可按照自身日常习惯完成活动，无需遵循标准化流程限制。原始数据首先会经过降噪与杂波抑制处理，该处理包含距离选择以及直流（DC）分量消除环节。随后，通过采用部分重叠的滑动时间窗口对数据进行分段，提取得到TR特征图。本数据集还提供了针对5名参与者的5折交叉验证实验所用的*.csv文件。