Solar and Kinetic Energy Harvesting Dataset for Human Activity Recognition (SKEH)

This data was collected from wearable accelerometer, solar and kinetic energy harvesters for human activity recognition. The data was collected to explore the performance of solar and kinetic energy harvesters in recognising various human activities to enable self-powered operation of wearable devices.\nLineage: Data Collection \nSensors Used:\nKinetic energy harvester: MIDÉ Technology S230-J1FR-1808XB two-layer piezoelectric bending transducer.\nSolar energy harvester: IXYS SLMD121H10L solar module\nAccelerometer: InvenSense MPU9250\nAll data stored at 100 Hz.\nData Collection Device (SEH and KEH): Beaglebone Green\nCollection Period: From November 2018 to June 2022\nEnvironment: Indoor and outdoor\nCollection Protocol: Data during 5 activities from each participant \n\n\nDataset Structure:\nFolder and File Organization: \nThe data is organised in 2 folders: \nIndoors\nOutdoors\n\nEach folder has three subfolders:\nAccelerometer\nKEH\nSEH\n\nFive files in each folder (running.csv, sitting.csv, stairs.csv, standing.csv, walking.csv)\nFile Formats: CSV for time-series data\nData Schema: \nKinetic energy harvester:\n`Data`: [Ampere]\nSolar energy harvester:\n`Data`: [Ampere]\nAccelerometer:\n`acc_x`: [g]\n`acc_y`: [g]\n`acc_z`: [g]\n\n\nRaw Data Status: This is raw data segmented into 5 activities (break period were removed).\nPreprocessing Steps:\nResampling: Solar and kinetic energy harvesting data was downsampled from 100 kHz to 100 Hz using linear interpolation.\nSegmentation: Data was segmented into 5 activities and break periods were removed.\n\n\nPlease cite the relevant work.\n1- Sandhu, Muhammad Moid, et al. "Fusedar: Energy-positive human activity recognition using kinetic and solar signal fusion." IEEE Sensors Journal 23.11 (2023): 12411-12426.\n2- Sandhu, Muhammad Moid, et al. Self-Powered Internet of Things: How Energy Harvesters Can Enable Energy-Positive Sensing, Processing, and Communication. Springer Nature, 2023.\n3- Sandhu, Muhammad Moid, et al. "SolAR: Energy positive human activity recognition using solar cells." 2021 IEEE International Conference on Pervasive Computing and Communications (PerCom). IEEE, 2021.\n

本数据集采集自可穿戴加速度计、太阳能能量采集器（Solar Energy Harvester，简称SEH）与动能能量采集器（Kinetic Energy Harvester，简称KEH），用于人体活动识别相关研究。本数据集的采集目的为探究太阳能与动能能量采集器在识别各类人体活动时的性能，以实现可穿戴设备的自供电运行。

数据溯源：数据采集
所用传感器：
- 动能能量采集器：MIDÉ Technology S230-J1FR-1808XB 双层压电弯曲换能器
- 太阳能能量采集器：IXYS SLMD121H10L 太阳能模块
- 加速度计：InvenSense MPU9250
所有数据均以100 Hz的采样率存储。

数据采集设备（SEH与KEH）：Beaglebone Green
采集周期：2018年11月至2022年6月
采集环境：室内与室外
采集协议：采集每位参与者5类活动期间的时序数据

数据集结构：
文件夹与文件组织形式：
数据集分为2个主文件夹：室内（Indoors）、室外（Outdoors）
每个主文件夹下设3个子文件夹：加速度计（Accelerometer）、动能能量采集器（KEH）、太阳能能量采集器（SEH）
每个子文件夹内包含5个文件：running.csv（跑步）、sitting.csv（坐姿）、stairs.csv（上下楼梯）、standing.csv（站立）、walking.csv（行走）

文件格式：时序数据采用CSV格式存储
数据字段说明：
- 动能能量采集器：字段`Data`，单位为安培（A）
- 太阳能能量采集器：字段`Data`，单位为安培（A）
- 加速度计：字段`acc_x`、`acc_y`、`acc_z`，单位均为重力加速度（g）

原始数据状态：本数据集为经过活动分段的原始时序数据，已剔除休息时段数据。

预处理步骤：
1. 重采样：采用线性插值法，将太阳能与动能能量采集器的原始采样率从100 kHz降采样至100 Hz
2. 分段处理：将采集到的原始数据按5类人体活动进行分段，并移除休息时段数据

请引用以下相关研究文献：
1. Sandhu, Muhammad Moid, et al. "Fusedar: Energy-positive human activity recognition using kinetic and solar signal fusion." IEEE Sensors Journal 23.11 (2023): 12411-12426.
2. Sandhu, Muhammad Moid, et al. Self-Powered Internet of Things: How Energy Harvesters Can Enable Energy-Positive Sensing, Processing, and Communication. Springer Nature, 2023.
3. Sandhu, Muhammad Moid, et al. "SolAR: Energy positive human activity recognition using solar cells." 2021 IEEE International Conference on Pervasive Computing and Communications (PerCom). IEEE, 2021.