Dark Mode
Elderly Activity Recognition Dataset
Data Science and Analytics
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This dataset, known as Human Activity Recognition 70+ (HAR70+), is a professionally-annotated collection of data designed for activity recognition in senior citizens. It features recordings from 18 older adult subjects, ranging from fit to frail, aged between 70 and 95 years old. Each subject wore two 3-axial accelerometers, one on the right thigh and another on the lower back, for approximately 40 minutes during a semi-structured, free-living protocol. The dataset provides valuable insights into the daily movements and activities of senior individuals, supporting research and development in health monitoring and assistive technologies.
Columns
The dataset is structured with the following columns:
- timestamp: Records the date and time when each sample was collected.
- back_x: Acceleration data from the back sensor in the x-direction (down), measured in units of 'g'.
- back_y: Acceleration data from the back sensor in the y-direction (left), measured in units of 'g'.
- back_z: Acceleration data from the back sensor in the z-direction (forward), measured in units of 'g'.
- thigh_x: Acceleration data from the thigh sensor in the x-direction (down), measured in units of 'g'.
- thigh_y: Acceleration data from the thigh sensor in the y-direction (right), measured in units of 'g'.
- thigh_z: Acceleration data from the thigh sensor in the z-direction (backward), measured in units of 'g'.
- label: An annotated activity code corresponding to a specific activity. The coding scheme is as follows:
- 1: walking
- 3: shuffling
- 4: stairs (ascending)
- 5: stairs (descending)
- 6: standing
- 7: sitting
- 8: lying
Distribution
The dataset is provided in individual .csv files for each subject's recordings. A sample file, such as '501.csv', is approximately 9.36 MB in size and contains 8 columns. Across the entire dataset, there are around 104,000 records in total, with all columns exhibiting 100% validity and no missing values. The timestamps are consistently from 24th March 2021.
Usage
This dataset is ideal for various applications, including:
- Developing and evaluating human activity recognition (HAR) models for older adults.
- Research in gerontology and assistive living technologies.
- Machine learning tasks such as classification and time series analysis.
- Health and fitness monitoring of elderly individuals.
- Educational purposes in computer science and data science curricula focusing on sensor data processing.
Coverage
The dataset's demographic scope includes 18 older-adult subjects who are described as fit-to-frail, with ages ranging from 70 to 95 years old. The time range for data collection is specific to 24th March 2021. There is no specific geographic scope mentioned.
License
Attribution 4.0 International (CC BY 4.0)
Who Can Use It
This dataset is particularly useful for:
- Researchers in fields such as computer science, gerontology, and bioengineering interested in human activity recognition and sensor-based health monitoring.
- Data scientists and machine learning engineers looking for real-world sensor data to train and test classification and time series models.
- Healthcare technology developers aiming to create solutions for elderly care, fall detection, or activity tracking.
- Academic institutions and students studying signal processing, embedded systems, or applied machine learning.
Dataset Name Suggestions
- HAR70+ Senior Activity Data
- Elderly Activity Recognition Dataset
- Geriatric Mobility Sensor Data
- Senior Citizen HAR Dataset
- Activity Patterns of Older Adults
Attributes
Original Data Source: Elderly Activity Recognition Dataset
QUALITY