mAid

A mobile solution of timely and effective health aid for elderly living alone

The world’s population is ageing, with around 13% of the global population at ages 60 and above in 2017, and by 2050 this proportion will rise to nearly 25% in all regions of the world except Africa.

On the one hand, aging population significantly increase the demand for disease management, long-term care and better trained workforce at the community level; and also create physical, emotional as well as fiscal burden at the household level. On the other hand, many older persons want to remain active and independent life style, and they tend to not report some physical injuries, like falls, to the families or even doctors, although falling is the number one cause of fatal and nonfatal injuries among seniors. Another major concern of older people living alone is their inability to ask for help when they are paralyzed or unconscious.

To prevent fatal consequence of sudden conditions and support self-management for elderly living alone, we propose a mobile solution, mAid, to support timely and effective health aid through Android-based mobile devices. It leverages wearable computing and machine learning techniques to provide real-time fall detection, as well as emergency intervention under certain circumstances. When a risky event is detected during the continuous activity monitoring, mAid first interacts with the user to confirm his/her current physical condition; if the user is unable to response, mAid will notice the designated person or healthcare provider for emergency aid and treatment. 

Instead of introducing another specific manufactured apparatus, the proposed solution is an application running on Android operation system, and hence, it has great generalizability due to the popularization of Android-based mobile devices. In addition, many fall detection products on the market allow the user to call for help by pressing a button on the body-worn device, however, it is not effective when the user is not able to talk or in a state of unconsciousness. Because the proposed solution does not need user to take initiative and ask for help, it works autonomously to support timely and effective health aid in dynamic situations.

This solution is a mobile application to support health aid using sensor technology and advanced computational algorithms. It is innovative due to the generalizability and autonomy. Most existing commercial products with similar goals have their own manufactured equipment, and they allow the user to call for help by pressing a button on the body-worn device. On the contrary, the proposed solution leverages the popularization of Android-based mobile devices, and takes initiative to act once the fall has been detected. Therefore, this solution does not tie up with any specific device, and it works effectively in dynamic situations.

Technology is integral to the proposed solution in four ways: (1) sensor signal processing is used for long-term continuous physical activity monitoring; (2) feature engineering and machine learning is used for real-time event detection; (3) transfer learning is used for automatic model adaptation across diverse data sources; and (4) mobile technology is used for user intervention and emergency alert.

2018-08 to 2018-10: Complete the prototype of mAid app on smart watches and test major functionalities.

2018-10 to 2019-02: Find partners (healthcare provider or relevant research group) to test the app on individuals (preferably older people) in everyday living environment, and improve the performance according to the feedback.

2019-02 to 2019-03: Summarize the test results (precision, recall, response time, energy consumption, etc.) and compare against similar products, to enhance current collaboration and find new partners for more impact. 

2019-03 to 2019-07: Release the app for public use, and maintain a webpage for discussion and feedback.

There are two major approaches to grow and scale the proposed solution: (1) along with the development and commercialization of wearable sensing devices, the current solution will be extended to support other types of sensor-based platform such as chest-worn or head-worn equipment, to serve a wide range of usage; and (2) in addition to basic functionalities built in the app, customized motion analysis service will be offered to healthcare providers and other kinds of partner for profit.

The proposed solution targets two types of customer, individuals in need and healthcare providers. For the former, once mAid app is available to download online (within the next 12 months), they can install and use the major functions for free. For the latter, I plan to reach out more audiences through the event like SOLVE, and (1) directly contact healthcare providers to provide long-term remote service to older patients living alone or in rehabilitation; (2) build connection with relevant research teams (both academic and industrial), to collaborate on advanced motion analysis research and deploy the solution through their channels.

The proposed solution is currently in prototyping, and it will be available for test in 3 months.

Due to the light-weight and personalized design of this application, it can work independently on the end device (i.e. smart watch) with no requirement for centralized data storage or process. The communication for health aid is carried through the existing cellular network. Therefore, the growth of customer population will not be an issue of service quality.

I’ve been working as a research assistant during my PhD program for four years, and my research focus is wearable computing and machine learning for smart health applications. I’ve participated in two clinical studies collaborated with medical school and children hospital for sensor-based mobility assessment, and worked on three projects funded by NSF and NIH, to design and develop scalable machine learning frameworks for activity recognition. I believe my experience of clinic study and system design can contribute to the success of this solution.

The revenue model of this solution is the freemium model, in which the basic functionalities are free for individual users, yet the organizations (i.e. healthcare provider) need to pay for customized analysis services and extensions.

I’d like to get insightful feedback and suggestions from domain experts to advance the current solution, and approach to more audiences through the Solve challenge. In the years of working as a research assistant, I realized it is more difficult yet important to ask the right questions than to find the right answer. Although I’m passionate about the technology itself, I want to make sure it is applied to solve an unmet problem in an effective way. To that end, Solve can help me to refine my plan and stay on the right track.

The key barrier of my solution is how to attract organization partners and customers. Solve can help me to build connection with relevant research groups and healthcare providers, so that I can find opportunities for future collaboration.