AIOT Design & UX Design
FitPal is an iOS app meant to enhance users’ fitness level, improve user’s fitness efficiency and provide them a pleasant experience with adaptive playback adjustment, movement monitoring,and haptic feedback. Fitpal keeps fit and healthy with your friends and families.
4 weeks (2021.11-2022.12)
Adviser: Yuntao Wang
Apple watch, iphone,
User data collector
User availability tester
who need fitness guidance
A Watch App with iOS app as companion
·Recognize and monitor user movements’ type and exercise data
·Provide fitness guidance through vibration and other visual feedbacks
·Improve user’s fitness efficiency
·Provide them pleasant experience
THE VIDEO OF FITPAL
Video duration: 2'11''
We collected 73 questionnaires, conducted 2 focus interview, and carried out 2 contextual Inquiry to found that many people used Watch for health monitor but few people used it for fitness guidance, which indicated that there was chance for us to design new interaction ways to guide users when training.
Synthesizing these data and previous research, we produced Journey maps, Persona, etc. to identify our users and find out their problems and needs.
·DEFINE USER'S NEED
According to the classification of user needs, We generated user persona to help us identify potential user needs and their end goals.
·JOURNEY MAP & OPPORTUNITY
To sum up, the above researches helped us to have a sense of our users’ problems, what they want to achieve and opportunities we can do to help them, that is Using Apple Watch to record motion data from user, and count how many actions did user did, automatically changing the speed of the tutorial videos, and reducing the need to watch the screen, thus more convenient.
·IOS UI Hi-Fi DESIGN
FitPal provides you with the functions: frequency adaption, quick start, movement recognition, specific & communicative vibration feedback, and breathing adaption by using simple, highlighted, pleased user Interface & its interaction with few steps.
·Adjust the speed of the tuition video to user’s own speed.
·After clicking "Start", you can start exercising with the system automatically recognizes your movements and matches the course for you, and once recognizing that your movements are not standard, FitPal’ll stop all functions and give visual feedback.
DATA COLLECTION & ALGORITHM
At first, we choose 5 types of movement to develope.
We first judge every single data point and their previous point.
·Decide whether the data point has a potential to rise rapidly.
·Find out where do the data stream start to rise rapidly.
·Find out whether the previous n2 data of this data is all 0, which means this data in this point is the first one to rise rapidly.
In this way we can find out the start point of each data stream.
And the data between two start point is the data of each movement.
·THE RESULT OF MODEL TESTING
Storyboard: Sketch to Code
Each view extends WKInterfaceController (WatchKit)
HealthKit for both iOS and WatchKit Extension
Translate from MATLAB to Swift
Called every one second to pass on the real-time data
FireBase (for user authentication)
AVKit / AVFoundation (for video player)
👈 Part of the code
·VISUAL IDENTITY-LOGO & ICON & FONT & COLOR
We designed a Usability Test to measure our mobile phone and watch app. And we had three main evaluation process, Pre-Test Questionnaire, User Tasks, and Post-Test Questionnaire. The questions raised therein are mainly aimed at accuracy of our algorithm, user experience and users’ advice.
The user experience value of this time is higher than the user experience of other software such as Keep that we collected before.
In these four indicators, the average user experience is above the median, but in the experience of vibration feedback, the user is below the positive number 5.
There was too little distinction between the two vibrations, such as telling the user to take a break after a set，like strong/different vibration feedback for resting.
·USER EXPERIENCE FEEDBACK
IN THE FUTURE
·DEVELOPMENT & ALGORITHM
Update The Algorithm
Accuracy needs to be improved
Advanced ML/Deep Learning method
Change segmentation method
Bug Fixing & Optimizing
Collapse & Flashback
Consuming Too much RAM
·POTENTIAL PATCH DESIGN
We found that the error in the final judgment was not only a problem of model training, but also a lack of more accurate positioning, so it’s a better way to design a series of patch products both with 5 primary patches and at least 4 patches to attach to users’ body, in order to providing us more accurate relative coordinates.
·Cancel Video adaptation - based on
·Countdown Voice feedback
·Check for adequate warm-up
·Voice feedback for counting and finishing
·Specifications is needed
·Then strong/different vibration feedback for resting
·sensor can be added to other parts of our body