A mobile application to assist Alzheimer’s patients and caregivers

Keh, Yi Qian (2025) A mobile application to assist Alzheimer’s patients and caregivers. Final Year Project, UTAR.

PDF Download (12Mb)

Abstract

Alzheimer's disease (AD) is a progressive neurological disorder that severely impacts memory, cognitive functions, and daily living abilities, posing significant challenges not only for patients but also for caregivers and society. With no known cure, there is a critical need for supportive solutions to alleviate these difficulties. This project aims to develop a mobile application that assists Alzheimer's patients and their caregivers. Therefore, MemoraCare was developed, and it is a Flutter-based mobile application that helps caregivers and patients stay connected, safe, and recognized in real time by integrating functionality such as real-time face detection and recognition, safe-distance navigation, communication channels, AI assistance, diaries, and NFC info cards, with Firebase as a backend service. There are three models, which are the custom CNN model, Siamese Network, and MobileFaceNet.. The fine-tuned MobileFaceNet achieved 80.5% testing accuracy, which clearly outperformed the other models and emerged as the most suitable for this project. Compared to the custom CNN, which suffered from overfitting, and the Siamese network yielded modest results. Firstly, the app delivers an end-to-end, on-device face pipeline designed specifically for caregiver–patient workflows: a lightweight embedding-based recognizer by using the MobileFaceNet to enable open-set identification, indicating that new people can be added without retraining, while the ML Kit is used for face detection. Secondly, by coupling recognition with a person-linked memory repository, MemoraCare enables the recall of the patient's memory. Thirdly, the app provides a safety stack combining geolocation by allowing the caregiver to set a customized safe zone, Haversine distance for geofencing to calculate the shortest distance between the patients and caregiver, and walking-route retrieval to guide the patient return to the designated locations for real-world caregiving scenarios by integrating the Google Cloud APIs: Maps API, Directions API, and Places API to display the locations on the Google Maps, suggest a route between two points, and search for places. Fourthly, the app increases the relationship between the patient and caregivers by integrating a communication chat room that employs deterministic pairwise chat IDs and voice-note peak compaction for achieving 1 to 1 chat room for the users and supports voice messaging. Also, a role-conditioned OpenAI assistant that is driven by carefully designed prompts to respond v Bachelor of Computer Science (Honours) Faculty of Information and Communication Technology (Kampar Campus), UTAR with an expected reply. Fifthly, the app writes a minimal public-profile URL to an NFC NTAG213 tag to enable tap-to-view access so bystanders can quickly contact caregivers if the patient becomes lost. Sixthly, the app encourages the patient to write and record a diary about their valuable and interesting memories by storing it in the cloud instead of their brain, so that they can review and recall the memories if needed. Hence, the application attempts to enhance daily living, promote memory recall, improve communication, ensure patient safety, and reduce caregiver burden.

Actions (login required)