watchHAR: A Smartwatch IMU dataset for Activities of Daily Living

# watchHAR: A Smartwatch IMU dataset for Activities of Daily Living

This is a dataset of IMU recordings (3-axial acceleration, 3-axial gyroscope, and 3-axial magnetometer) recorded with Sony Smartwatch 3, along with ground truth location data from a vicon motion capture system. Smartwatch IMU and vicon location data exist for the users' left and right hands, as well as vicon location data from their left and right ankles.

The recoded data cover the following `activity_id`s:
* `brushing_teeth`
* `idle`
* `preparing_sandwich`
* `reading_book`
* `typing`
* `using_phone`
* `using_remote_control`
* `walking_freely`
* `walking_holding_a_tray`
* `walking_with_handbag`
* `walking_with_hands_in_pockets`
* `walking_with_object_underarm`
* `washing_face_and_hands`
* `washing_mug`
* `washing_plate`
* `writing`,

all of which were recorded in the same room, across the span of several days. In addition to these activities, walking stairs up and down (`activity_id`s `stairs_up` and `stairs_down`) events were recorded in various different locations. Note that for these stairs events only Smartwatch IMU recordings exist, and only from the dominant hand of the participants, which coincided with the right hand in all cases.

The data are organised in the following folder structure:
`user_id/activity_id/recording_type.csv`. Stair events are a special case in that multiple up/down stair recordings can exist for a single user. These are separated like the following example
```
`user_01/stairs_down-02/smartwatch_right_hand.csv`
```
that points to the IMU data of the second `stairs_down` recording of the first user.

Finally, note that all timestamps refer to a UTC timezone, and that data collection took place in the United Kingdom.

# watchHAR：面向日常活动识别的智能手表IMU数据集
本数据集依托索尼智能手表3（Sony Smartwatch 3）采集惯性测量单元（Inertial Measurement Unit, IMU）数据，包含三轴加速度、三轴陀螺仪及三轴磁力计录制数据，同时配套Vicon运动捕捉系统采集的地面真值位置数据。智能手表IMU数据与Vicon位置数据均覆盖受试者的左右手，此外Vicon位置数据还覆盖受试者的左右脚踝。

本次录制的数据涵盖以下`活动标识（activity_id）`：
* `brushing_teeth`：刷牙
* `idle`：空闲状态
* `preparing_sandwich`：制作三明治
* `reading_book`：阅读书籍
* `typing`：打字
* `using_phone`：使用手机
* `using_remote_control`：使用遥控器
* `walking_freely`：自由行走
* `walking_holding_a_tray`：端托盘行走
* `walking_with_handbag`：提手提包行走
* `walking_with_hands_in_pockets`：双手插兜行走
* `walking_with_object_underarm`：腋下夹物行走
* `washing_face_and_hands`：洗脸洗手
* `washing_mug`：清洗马克杯
* `washing_plate`：清洗餐盘
* `writing`：书写

所有上述活动均于同一房间内录制，采集周期横跨数日。除上述活动外，数据集还采集了上下楼梯（活动标识`stairs_up`与`stairs_down`）的数据，采集场景涵盖多处不同地点。需注意的是，此类楼梯活动仅采集了智能手表IMU数据，且仅来自受试者的利手——所有受试者的利手均为右手。

数据集采用如下文件夹组织结构：`user_id/activity_id/recording_type.csv`。楼梯活动数据属于特殊情况：单个受试者可对应多条上下楼梯的录制数据，此类数据的命名格式如下例所示：
`user_01/stairs_down-02/smartwatch_right_hand.csv`，该路径指向受试者01的第2次下楼梯录制的右手智能手表IMU数据。

最后需说明，所有时间戳均采用协调世界时（UTC）时区，且数据采集工作均在英国完成。