Dark Mode
Walker Activity and Fall Data
Data Science and Analytics
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
A curated compilation of inertial data focused on studying fall detection systems for individuals who use walking assistance devices. This data offers deep insight into various movement patterns, capturing four distinct classes: idle, motion, step, and fall. The data set was published as part of a research paper related to activity logging and motion classification for orthopedic walkers.
Columns
The dataset is presented in a CSV file containing 961 columns.
label: This is the classification column, covering the four movement categories (idle, step, movement, fall) intended to be predicted. It contains 4 unique values and is 100% valid across all 2480 records.
acc_x_0 to gyro_z_159: These 960 feature columns encapsulate processed sensor readings. They represent 160 samples each for acceleration and gyroscope data across the x, y, and z axes. These are functional representations of the time series data.
Distribution
The data is provided in a single CSV file,
full_dataset.csv, which is approximately 16.18 MB in size. The collection consists of 2480 samples in total. The classes are evenly balanced, with each class containing 620 individual samples.The raw data was originally recorded using an Arduino Nano 33 BLE Sense IMU affixed to a walker. Data was processed by reducing the rate to 80 samples per second to ensure a constant time step and synchronize acceleration and gyroscope readings. Segmentation, using algorithms like the Hidden Markov Model, was performed to extract steps and falls, resulting in samples windows of 160 readings (equivalent to a 2-second duration).
Usage
- Developing machine learning models to classify walker activity, such as distinguishing walking from idle periods.
- Designing and testing effective fall detection and alerting systems for mobility devices.
- Signal processing and feature engineering exercises using processed time-series inertial sensor data.
- Academic research into the dynamics of movement involving orthopedic walkers.
Coverage
The data captures inertial readings from an IMU device attached to a walker. Data collection involved four different subjects maneuvering the walker down a hallway, primarily capturing step and movement patterns. The 'idle' data represents periods when the walker was stationary and is not subject-specific. Similarly, 'fall' data was generated by deliberately pushing the walker over and is not linked to any particular individual. The core data captured are six dimensions of acceleration and gyroscope data.
License
CC0: Public Domain
Who Can Use It
- Engineers and Developers: Working on real-time embedded systems or mobile applications for monitoring patient safety.
- Data Scientists: Utilising classification algorithms like Random Forest for activity recognition.
- Students and Researchers: Focused on signal processing, sensor fusion, and mobility health technologies.
Dataset Name Suggestions
- Walker Activity and Fall Data
- Inertial Walker Motion Sensor Log
- Assisted Mobility Device Classification
- IMU Data for Orthopedic Walker
Attributes
Original Data Source: Walker Activity and Fall Data
QUALITYLoading...