IPqM-Fall

The IPqM Fall Dataset was created as part of the "Future Combatant" project by the Navy Research Institute (IPqM), with support from Federal Center for Technological Education Celso Suckow da Fonseca(CEFET-RJ) and funding from Funding Agency for Studies and Projects (Finep). The project's objective is to classify falls occurring during military operations, distinguishing them from other activities. To achieve this, it is necessary to collect data from daily life activities and military activities, including the use of weapons. This approach makes the dataset useful for both military and civilian contexts, as it includes activities such as walking, sitting, running, jumping, among others.

For data collection, wearable devices were used, specifically the Samsung Galaxy Watch 4 smartwatch and the LG Velvet smartphone. The collected variables include:

Linear Acceleration: Captured by the accelerometer, measuring the rate of change in linear velocity along the x, y, and z axes.

Angular Acceleration: Captured by the gyroscope, measuring changes in angular velocity along the x, y, and z axes.

Data was collected from 15 military personnel from the Navy Research Institute, with different ranks. Each volunteer used an Android smartphone in the chest pocket and two WearOS smartwatches on their wrists. Participants had to be qualified for armed service and have at least one year of experience. Detailed information about the participants, including rank, gender, age, height, and weight, was recorded to ensure a representative sample of the military population.

Three categories of activities were collected:

Activities of Daily Living (ADL): Basic tasks such as standing, walking, running, jumping, and sitting. The activities of daily living (ADL) collected were as follows: 1) standing, 2) walking, 3) running, 4) jumping, 5) sitting in a chair, 6) standing up from a chair, 7) walking uphill, 8) walking downhill, 9) running uphill, 10) running downhill, 11) walking upstairs, 12) walking downstairs, 13) jumping upstairs one step. The activities ADL_1, ADL_4, ADL_5, ADL_6, ADL_11, ADL_12, ADL_13, and ADL_14 were also performed while carrying a rifle, considered as military operation.

Military Operations (MO): Tasks performed during military operations, including different shooting positions and crawling. The collected activities were as follows: 1) Sweep with walking, 2) Sweep with quick engagement, 3) Transition from standing to kneeling shooting position, 4) Transition from walking to kneeling shooting position, 5) Transition from running to kneeling shooting position, 6) Transition from standing to prone shooting position, 7) Transition from walking to prone shooting position, 8) Transition from running to prone shooting position, 9) Crawl.

Fall Activities: Simulated falls in various directions. The activities collected were: 1) forward fall ending lying on the back, 2) forward fall ending lying face down, 3) backward fall, 4) right side fall, 5) left side fall.

The dataset is organized into directories containing subdirectories for each participant ID. Each ID subdirectory contains subdirectories for smartphone (chest) and smartwatch (left and right wrists) data. The files include:

sampling.csv: Contains activity IDs, sensor positions, timestamps, and rifle usage indication.

acceleration.csv and angular_speed.csv: Contain the magnitude and the x, y, and z components with timestamps of each observation.

This comprehensive dataset provides information for training neural networks and other machine learning techniques, allowing the analysis of various activities and potential fall events in both civilian and military contexts.