Detection and Classification of Activities of Daily Living Through EEG

Project Title

Detection and Classification of Activities of Daily Living Through EEG

Project Area of Specialization NeuroTechProject Summary

Current decade witnessed significant research in the detection of Activity of Daily Living (ADL) using multiple wearable sensors. The ADL are the routine tasks that an individual does independently. Human Activity Recognition (HAR) aid vastly in understanding human behavior by detecting human activity with the help of multiple sensors. This monitoring of the ADLs can help in deciding the medical status of the person being monitored. Failure to do these tasks may indicate presence of a medical condition.

HAR is a very powerful tool in all sectors of the modern living. This concept can be used to monitor and control different aspects of industry, as well as, households. In the modern era, HAR, alongside with IoT, is being used to monitor the activities of people in smart homes to keep a track of their safety, and to provide surveillance & security. This concept is also being implemented in smart offices to detect the human activity to achieve maximum power and energy control.

The project proposes a non-invasive method to monitor the Activities of Daily Living, classify them and to detect any type of abnormal activity with the help of the results obtained by EEG of patients. EEG can be used to recognize human activity at any time of the day. EEG provides different signals for different activities, making it easier to detect any kind of abnormality in the person from the abnormal signals obtained. Different medical devices can be used to obtain the EEG of the people in their daily lives. Recently a wearable device used to gain the EEG signals is Neurosky Mindwave Mobile 2 ®.

https://www.dropbox.com/s/586snr8vb7w4e02/lucid chart.png?dl=0

Project Objectives

The project uses the idea of smart health-care systems and IoT to develop an application that would not only monitor & classify ADLs, but also detect abnormal activity and alert the caretaker of the patient in case of emergency. This monitoring of the subject is done using wearable and portable EEG sensor which record the data of the subject during different activities. Once the data has been collected, statistical analysis is done on it for enhanced accuracy. The data is processed and classified in the mobile phone from where the categorized data is transmitted to the cloud. Appropriate pattern recognition framework at the cloud is employed to make the decision of alerting the caretaker by classifying emergency situations. The system is to be tested on different people to confirm its accuracy. Furthermore, the developed system is also compared against the already existing accelerometry system to verify the improved accuracy which is expected using EEG.

The main objectives are:

• Selection of wearable sensors to obtain the EEG scans of the user for monitoring and detection of ADLs.
• Apply machine learning algorithm that understands the data obtained from the EEG scans and makes decision based on the real data fed to it beforehand.
• Creation of interactive mobile application.
• Database handling, classification and accuracy.
• Comparing the results of detection and monitoring of these EEG scans against the results of the already existing accelerometer-based system.

Project Implementation Method

The data collection is done using the portable EEG device Neurosky Mindwave Mobile 2.0. In order to collect the data, we connected the portable EEG device to our laptops via Bluetooth. The data was collected from different subjects at different times of the day, while they were performing different activities. These activities are classified into 4 major classes:

• Stationary class (sitting, sleeping, standing, etc.)
• Light Ambulatory class (walking, climbing stairs, cooking, etc.)
• Intense Ambulatory class (running, jogging, jumping, etc.)
• Abnormal class (losing consciousness, falling, etc.)

After the datasets are collected and classified into one of these categories mentioned above, statistical analysis is done in time and frequency domain using MATLAB. The datasets are then fed for model training by applying different classifiers using WEKA. Once the model training is complete, it is tested for accuracy using test sets. Upon achieving the desired accuracy, a mobile phone application is created, which is integrated with our EEG device. Moreover, a mobile database (google firebase) is used to save all the data of the user. Once the classified data is sent to the database, the decision of whether to call the caretaker in emergency situations is be made.

The developed application is tested and compared with the already existing application that uses accelerometry to confirm the enhanced accuracy.

Benefits of the Project

The benefits include:

• This can be very beneficial in the health sector as it can be used to identify whether something is wrong with the personby observing their daily living activities. 
• The application will be able to detect the mood of the subject from their EEG signals. In case of any depressive mood, it will suggest different activities to the subject. Hence, we may integrate it with existing social media applications such as food panda, uber, facebook, maps, easytickets and calendar applications etc. for different suggestions, to improve the mood of the subject.
• This application will provide round-the-clock monitoring for those in need. This means that the doctors will be able to properly monitor the patient, without them having to stay in the hospital environment. 
• The proposed project will provide enhanced productivity and common wellbeing. If the people are healthy, they will be much more productive. Furthermore, due to the constant round-the-clock monitoring, they will not need a lot of healthcare.

Technical Details of Final Deliverable

The hardware we have used to obtain the EEG signals is a sensor known as Neurosky Mindwave Mobile 2 ® which is connected through Bluetooth with the smartphone in which the application is installed. This sensor takes the EEG signals of the subject’s brain that are further processed to identify the activity performed, as different activities give different EEG signals. The signal is recorder for a specific time duration and it is transmitted from the sensor to the mobile application where it is classified into one of the four classes of ADLs. If the activity is classified as a normal class then it is simply stored in the cloud based database but if the activity is classified as an abnormal class then along with storing in the database, the caregiver of the subject is also notified by a SMS along with the location of the subject. Mobile GPS defines the location of the subject in need.  The subject is also prompted in case of abnormal behaviour so if there was incorrect detection, we can prevent giving trouble to the concerned people. The developed mobile application is userfriendly and can be used by both suject and caregivers. Other hardware used is IMU and Arduino to record the accelerometer data by which we compared the accuracies with already existing system. The user has to register in the mobile application and once the account is created, the user is assigned a unique ID against which his data is stored in the database and the history of his activities is maintained.  

Final Deliverable of the Project HW/SW integrated systemCore Industry MedicalOther Industries IT , Health , Telecommunication Core Technology NeuroTechOther Technologies Artificial Intelligence(AI), Internet of Things (IoT), Cloud Infrastructure, Wearables and ImplantablesSustainable Development Goals Good Health and Well-Being for People, Decent Work and Economic Growth, Reduced InequalityRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
Neurosky Mindwave Mobile 2.0	Equipment	1	40000	40000
Mi Band	Equipment	3	7000	21000
IMU	Equipment	2	650	1300
Arduino UNO	Equipment	2	1200	2400
Arduino Pro Mini	Equipment	2	1000	2000
Wires, Male to Female, Male to Male, Female to Female	Equipment	2	50	100
Bluetooth Module HM-10 or equivalent	Equipment	3	800	2400
Miscellaneous, mobile app UI/UX development etc.	Miscellaneous	1	10000	10000
breadboard	Equipment	2	400	800