Lowng

User Dashboard
Infrastructure Diagram

Inspiration

Loneliness is a pervasive issue in America which impacts the elderly particularly hard, especially for Asian Americans, this is especially important when considering 30–40% of the residents in SF Chinatown are 60. Many Asian elders also face additional challenges such as language barriers, cultural differences, stigmas of mental health, the shift from intergenerational housing to single residency occupancy (SRO) housing. These issues are on top of the already apparent issues such as technological barriers, or family moving to work in other areas in a more interconnected economy.

What it does

We take an EEG headband, a device that can track the brainwaves being emitted by the brain and turn them into user insights about loneliness or stress. It then uses machine learning to send alerts or notifications to caretakers or family to call in or set a reminder to visit. This is done to promote social interactions and rebuild family bonds. The app also offers an overview to the user about their own emotional state and well-being.

How we built it

From the Muse S gen 2 that only gives us the 4 electrodes we convert those into Alpha Beta Theta etc. waves. From those waves I constructed a CNN-RNN fusion model that will give me the emotion of the user. To determine the boredom score we use an XGBoost ML model to derive the accurate percentage. Finally to derive the stress score we use a mathematical algorithm which ratios the alpha and beta waves. This categorized/analyzed data is fed to a react and nextjs frontend for both the user and caregiver to utilize.

Challenges we ran into

One of the biggest challenges that we ran into was end-to-end integration due to a difference between preferred software stacks. Akhil was used to querying document databases, while Brandon primarily works with relational ones. We chose to compromise by directly querying JSON data from Savir's backend server. This related to the second biggest challenge in which the backend server has to run locally to connect to the EEG and run the CNN, making it difficult to access remotely from a deployed instance of the web app. In the end, we decided to demo the app locally to be able to access the backend server, although it is designed to be deployed on Cloudflare.

What's next for Lowng

If this project would be taken further, the events feature would be further developed to show local events and centers where games are played such as Mahjong. This would be done to promote a sense of community and encourage social interactions.

UI design would also be more unified across all all interfaces to match a certain theme.

Launch would initially be set in SF Chinatown, could expand to other locations with high senior populations. If proven successful, global expansion is also valuable opportunity especially in aging nations such as Japan and Korea.

If monetization were to be considered here are possible avenues: Subscription Service, Anonymized Healthcare Data (Research, Colleges, Hospitals etc.), Featured/Sponsored Events in the events tab, custom engineered specialized EEG headbands.