Wild-SHARD: Samrtphone Sensor-based Human Activity Recognition Dataset in Wild

Wild-SHARD presents a novel Human Activity Recognition (HAR) dataset collected in an uncontrolled, real-world (wild) environment to address the limitations of existing datasets, which often need more non-simulated data. Our dataset comprises a time series of Activities of Daily Living (ADLs) captured using multiple smartphone models such as Samsung Galaxy F62, Samsung Galaxy A30s, Poco X2, One Plus 9 Pro and many more. These devices enhance data variability and robustness with their varied sensor manufacturers.The sensor module, consisting of accelerometers and gyroscopes, was mounted in the front pockets (vertically, phone earpiece side up) of 470 adult subjects of diverse ages, genders, weights, and heights. Subjects performed activities naturally to capture authentic ADL data, including sitting, walking, standing, running, and navigating stairs indoors and outdoors.The dataset was collected at a sampling frequency of 100 Hz, covering six ADLs: sitting, walking, standing, running, going upstairs, and going downstairs. The attributes recorded in the dataset include acceleration due to gravity, linear acceleration, gravity, rotational rate, rotational vector, and the cosine of the rotational vector. This comprehensive dataset structure, as detailed in the provided equations, aims to improve real-time activity recognition and overall system performance by offering high-quality, realistic sensor data.

Wild-SHARD 是一款新颖的人类活动识别（Human Activity Recognition, HAR）数据集，其采集自无约束的真实野外场景，旨在弥补现有数据集普遍缺乏真实非合成数据的局限性。本数据集包含多组日常生活活动（Activities of Daily Living, ADLs）的时间序列数据，采集设备涵盖多款智能手机机型，例如三星Galaxy F62、三星Galaxy A30s、Poco X2、一加9 Pro等。不同厂商搭载的传感器模组提升了数据的变异性与鲁棒性。传感器模组由加速度计与陀螺仪组成，被佩戴于470名年龄、性别、体重、身高均存在差异的成年受试者的前裤口袋中，且手机保持听筒端朝上的竖直放置状态。受试者自然完成各项活动以采集真实的日常生活活动数据，涵盖坐姿、行走、站立、跑步以及室内外楼梯通行。本数据集以100Hz的采样频率完成采集，共包含6类日常生活活动：坐姿、行走、站立、跑步、上楼以及下楼。数据集记录的属性包括重力加速度、线性加速度、重力分量、旋转角速度、旋转向量以及旋转向量的余弦值。正如文中所附公式详述的那样，这一全面的数据集结构旨在通过提供高质量、贴近真实场景的传感器数据，优化实时活动识别效果与整体系统性能。