Dataset of Mobile Sensors based Human Physical Activities Recognition to Pandemic Spread Minimization (KAU-COVID19-AR-Dataset)

• This repo describes a data collection campaign and a set of resulting data obtained from smartphone sensors that show useful functions in reducing the spread of COVID-19 (epidemic).
• Database is released as a collection of CSV files containing more than 37K data samples, where each sample is composed of 7 features related to many mobile sensors, including accelerometer, gyroscope, speed, and GPS sensors, with the help of our application that we developed using Flutter application development platform.
• First we collect the each sensor data locally in mobile device in a CSV format and then send this that to the PostgreSQL server.
• In addition, a sample of each data is associated with a basic truth label that describes the user’s activity and the situation in which he or she was involved during a sensory examination (e.g., handshakes, hand washing, hand sanitizing and eyes nose and mouth touching, etc.).
• To avoid introducing any bias during data collection, we performed a field-based sensory examination, that is, using multiple companies mobile devices, and without identifying any barriers to user behavior when performing any activity.
• The set of data collected represents a useful real data source to describe and evaluate a comprehensive set of novel activities aimed at helping to reduce the spread of COVID-19.

Table shows the Dataset (KAU-COVID19-AR-Dataset): Activities, No of samples

Activities	Number of Samples
Walking	15866
Hand Washing	6422
Standing	5665
Sitting	2869
Hand Sanitizing	2105
Nose Eyes Touching	1770
Hand Shake	1397
Drinking Water	1404

Data Specification

Title	Discription
Subject Area	Artificial Intelligence, Machine Learning, and Deep Learning
Specific Subject Area	Smart-phone based human physical activities detection using machine learning and deep learning
Type of Data	Sensor’s Data including Accelerometer, Gyroscope, Speed and GPS Sensor.
Data Formate	Raw and Featured
Experimental Factors	Data was collected by performing multiple activities e.g. Hand washing, Hand sanitizing, Nose-eyes-mouth touching, and Hand shaking etc
Related Research Article	Abdul Wasay Sardar, Computers in Biology and Medicine, https://doi.org/10.1016/j.compbiomed.2022.105662 [6]

Value of the data

• This data is collected by performing multiple activities e.g. washing sanitizing and shaking hands, walking, sitting, standing, and drinking water, can be used to build pandemic spread minimization system.
• The data can be used for behaviour analysis of individual, and group of peoples, and also helpful for the healthcare monitoring, health surveillance, medical and disability assistance.
• The structure and quantity of data can be used for various activity detection tasks such as classification, recognition, and prediction.
• This dataset is - as far as we know - the first publicly available physical activities dataset for the pandemic spread minimization.

Dataset Description

• The novel dataset (KAU−COVID19−AR−dataset) collects by performing multiple activities such as washing hands, hand sanitizing, shaking hands, touching the nose eyes or mouth, sitting, standing, walking and drinking water.
• The location of the smartphone is very important for the acquisition and higher accuracy of the dataset. Place your smartphone in a compatible watch-like position. Figure 1 shows the placement of the position on the smartphone.
• The location of the smartphone is similar to the location of Figure 1: Position of Mobile phone is (arm Position) for activity data collection the smartwatch that has achieved market penetration of in the past [4].

Figure 1: Position of Mobile phone is (arm Position) for activity data collection

• It is important to note that the position of the smartphone on the participant’s body is fixed.
• For in these experiments, we used smartphones from various companies.
• The orientation of the smartphone was vertically along the forearm.
• Data for all activities was recorded at 10 samples per second. This sampling rate (10 samples per seconds) is sufficient to detect human physical activity [3].
• Figure 2: Dataset (KAU−COVID19−AR−Dataset): Each activity In addition, recent advances have shown that less than 10 samples per second is insufficient for activity detection [1][2].

Figure 2: Dataset (KAU-COVID19-AR-Dataset):each activity

• Accelerometer, gyroscope, and speed sensor data are acquired at a speed of 10 samples per second, and GPSsensor data data are acquired at a speed of 1 sample per seconds [5].

Note: If you use this dataset, you have to cite the paper [6].

References

[1] Shoaib, Muhammad, Stephan Bosch, Ozlem Durmaz Incel, Hans Scholten, and Paul JM Havinga. "Fusion of smartphone motion sensors for physical activity recognition." Sensors 14, no. 6 (2014): 10146-10176.

[2] Wu, Wanmin, Sanjoy Dasgupta, Ernesto E. Ramirez, Carlyn Peterson, and Gregory J. Norman. "Classification accuracies of physical activities using smartphone motion sensors." Journal of medical Internet research 14, no. 5 (2012): e2208.

[3] Marusenkova, T. A. "An algorithm for detecting the minimum allowable accelerometer sample rate for tracing translational motion along a Bézier curve." Технічна інженерія 1 (85) (2020): 147-154.

[4] Samsung Galaxy S2. Available online: http://www.samsung.com/global/microsite/galaxys2/html/ (accessed on 14 May 2014)

[5] Nguyen, Khuong An, Zhiyuan Luo, and Chris Watkins. "Epidemic contact tracing with smartphone sensors." Journal of Location Based Services 14, no. 2 (2020): 92-128.

[6] Sardar, Abdul Wasay, Farman Ullah, Jamshid Bacha, Jebran Khan, Furqan Ali, and Sungchang Lee. "Mobile sensors based platform of Human Physical Activities Recognition for COVID-19 pandemic spread minimization." Computers in Biology and Medicine (2022): 105662.