Human Posture Dataset
收藏资源简介:
该数据集包含从Invensense的MEMS IMU传感器MPU-6050收集的加速度计传感器值,用于识别人类的6种姿势:站立、坐着、睡觉、跑步、向前弯曲和向后弯曲。数据集以单个csv文件形式存在,大小为3.2MB,包含44800个样本,每个样本有6列数据,分别来自胸部和腿部的传感器。
This dataset comprises accelerometer sensor values collected from the MPU-6050, a MEMS IMU sensor by Invensense, aimed at identifying six human postures: standing, sitting, sleeping, running, forward bending, and backward bending. The dataset is presented in a single CSV file, with a size of 3.2MB, containing 44,800 samples. Each sample consists of six columns of data, sourced from sensors located on the chest and legs.
数据集概述
数据集名称
- 名称: Accelerometer sensor dataset for Human Posture Recognition
数据集内容
- 传感器类型: Invensenses MEMS IMU sensor MPU-6050
- 安装位置: 左胸和右大腿各一个
- 包含姿势: 站立、坐、睡、跑、前屈、后屈
- 数据格式: 单个csv文件,大小3.2MB
- 数据样本总数: 44800
- 数据列数: 6列(Ax1, Ay1, Az1, Ax2, Ay2, Az2)+ 1列标签
- 标签对应姿势:
标签 姿势 0 睡 1 站立 2 坐 3 跑 4 前屈 5 后屈
数据收集
- 收集时间: 2018年春季
- 收集地点: College of Engineering, Pune
- 参与者: 3名学生
- 数据收集设备: NodeMCU(ESP8266)
- 数据传输方式: HTTP over TCP/IP
- 采样间隔: 200ms
数据特点
- 未校准或过滤: 包含原始噪声
- 姿势样本数:
姿势 样本数 睡 8329 站立 8358 坐 7542 跑 3617 前屈 11504 后屈 5450
使用建议
- 数据混洗: 建议对单一姿势的数据进行混洗以增强分类器的鲁棒性
- 数据不均等: 不同姿势和不同参与者的样本数不等,已去除异常值,保证参与者舒适度
数据下载
- 下载链接: 点击下载
引用信息
-
引用文献: H. Kale, P. Mandke, H. Mahajan and V. Deshpande, "Human Posture Recognition using Artificial Neural Networks," 2018 IEEE 8th International Advance Computing Conference (IACC), Greater Noida, India, 2018, pp. 272-278.
-
引用格式:
@INPROCEEDINGS{8692143, author={H. {Kale} and P. {Mandke} and H. {Mahajan} and V. {Deshpande}}, booktitle={2018 IEEE 8th International Advance Computing Conference (IACC)}, title={Human Posture Recognition using Artificial Neural Networks}, year={2018}, volume={}, number={}, pages={272-278}, doi={10.1109/IADCC.2018.8692143}, ISSN={2164-8263}, month={Dec}, }




