RSA project

Research

Proposal: How can AI provide routine nutritional assessment for elderly persons in an outpatient setting?

In order to understand what particualr areas of malnutrition within the elderly i could address, i conducted reaserch into all of the affecting areas, with the prominent causes were dementia and medicational influences.

My reserch showed me that: "More than 80% of patients with Alzheimer’s disease (AD) have eating and swallowing disturbances, leading to 14% of community-dwelling people with dementia at risk of malnutrition".

My medicational reserch showed that, "250 of the most commonly used drugs are known to alter sense of taste and smell or cause nausea and may therefore reduce appetite within the elderly"

In order to collect primary and secondary data for my targeted aged group, I decided to conduct an intervew with my Grandmother on what she felt would empower herself and her peers.

The technolodgy that i used to conduct the interview, was Google's latest transcripter, that automatically live transcripts voices into words. I then placed all of the key words from the transcript into a word cloud to gauge what key needs would be most prevalent, when designing a system.

Whilst conducitng the interview she felt her peers would benefit from a system that focuses on a dementia diagnosis, as dementia affects the nutritional wellbeing of the elderly, leading to further cognitive decline. Key words such as: Eat, System, Intelligence, Time and Nutrition came as the key needs identified. These are the words i took as that main influence for my system.

Proposal: What is my proposed intervention?

Below is my first concept sketch of the design of how the system named "Malnumentia" would operate with regards to the multiple components of the system.

After drawing a concept design for the system, i created Story Board to show case the processes that will occur when the system is put to use.

The user journey allows for a detailed replication of the process that elderly individual will have to go through in order to make the process work.

In order to plan out the process of development for the project, i created a Gantt chart to keep track on the progres my own timings and progess.

For the project i decided to use the Arduino NodeMcu (ESP8266 ESP-12E CH340G) WIFI edition. Due to it's Wifi functionality, it enables it to function with an Amazon Alexa and other Wifi connected devices.

Arduino Software (IDE) allowed us to easily upload unique line of code, thanks to Sam he was able to identify, help me develop code. The code is able to list four devices on the code, such as "Study Lamp" and "CFL", which will be able to be independently alter the settings via the Alexa, as shown later in the development of the project.

After developing the code, i uploaded it to the Arduino (as shown on the right). The Arduino confirms the upload via a blue led light on the front top base of circuit board.

In order enable a connection with the Amazon Alexa device, i had to connect the Arduino to the same local Wifi network, which was done by entering the wifi information as shown on the left.

After connecting the Wifi to both Alexa and the Arduino, i went onto the Alexa app and discoverved that the Alexa had indeed picked up the five virtual devices. This confirms the abilty to connect and create two devices together.

On the right it showcases the five devices, where i could individually control them from the app. From here i was able to connect them to an Alexa routine.

Proof of concept

Here i'm showcasing the abilty to connect verbal commands with the Amazon Alexa, to control the on/off state of the "CFL" device.

This highlights the connectivity ability for Alexa to function successfully via verbal commands, where its able to control the functions of the device.

These commands take the functionality of the device one step further, as im able to verbally increase or decrease the brightnees of the device, which as you can see only controls the on and off state of the device.

Theoretically the voice commands would be able to relay to an actual red light, controlling the brigtness of the device.

Further Development

Next i started looking into creating a custom Alexa skill that would remind the user on when to eat during the specific times of the day.
I coded a custom reminder into the application, with three key functions in the skill:

1. Frequency
2. Time
3. Purpose.

I began by signing up to Alexa's developers account, enabling me to build my owns custom alexa skills.

I then chose the correct model, being a custom choice.
Secondly i then decided what the correct host backend resource should be, with that being Node.Js and Java script

Once opening the package.Json file, i then Next i upgraded the ask-sdk software to the latest release 2.10.1

Next i added the lastest version of the moment timezone into the json script.

I then created an intent, where the Alexa would ask if i would like a reminder, with the basic response being 'Yes' .

On top of that, i created a custom utterance of 'sounds good!' which could be one of the user's responses to the intent.

Next i added the spoken reply that Alexa would ask to the user if they wanted a reminder to be set at 1pm, for eating lunch. This directly addresses the problem for users who suffer from dementia as mentioned previously.

In order to get alexa to create the reminder, i had to instruct it to create the basic intent of the reminder.

One step on from the above, i added the remider part of the Json script, where alexa would put forward a type of reminder for the application.

On line 31, i instructed alexa to push that notification onto the user's phone, where it would then sync with the user's calendar/ reminders app.

On line 49, i declared a reminder request obeject, with the relative being that it is offset from the current time. I changed the offset time to 10 seconds and turned on the push notfications on the phone.

On line 59, i wrote a line of text that Alexa would say if it had the infomation of the live stock content of the user's cupboards.

For example i said that it mentions that user has, pasta, egg and cheese and then asked if it would like to push for a recipie online. This would be true with the complete out of concept malnumentia model

Finally, i included the momenttimezone API on line 50 into the js code, as it allows for the user to set daily reminders by implementing the current date and time into the system.

To conclude, the concept of the development of the project, has allowed me to explore the potential with the interation between the user and the device, giving me an insight on how i could make malnumentia a viable system. I have been able to demonstrate a basic understanding of the way that the json and java script works and how they can be implemented into a prototype.