The Bite Counter
Adam Hoover, Eric Muth
Electrical & Computer Engineering Dept
Psychology Dept
Clemson University
Outline
1. The obesity problem
2. Our concept
3. How it works (and how well)
4. Manufacturing
5. Pilot study
The Obesity Problem
More than 27% of U.S. adults, ages 20-74, are obese
U.S Department of Health and Human Services, 2007
Worldwide, over 1 billion adults (ages 15+) are overweight or obese
World Health Organization, 2009
Related health risks include cardiovascular disease, diabetes,
cancer, and hypertension
$117 billion annualy to treat obesity-related issues in U.S.
Kelly, “Obesity: Health and Medical Issues Today”, 2000
New tools needed
International Obesity Society: 10,000 researchers
Obesity Society (U.S.): 2,500 researchers
• Joint NSF/NIH Workshop
•
(Ershow et al. 2007) “Engineering Approaches to Energy Balance and
Obesity: Opportunitites for Novel Collaborations and Research”
• Journal of the American Dietetic Association
•
•
3/21/2016
(Thompson et al. 2010) “Need for Technological Innovation in Dietary
Assessment”
(McCabe-Sellers et al. 2010) “Advancing the Art and Science of Dietary
Assessment through Technology”
4
Balancing Energy Output and Intake
exercise
X
diet
+
Y
healthy weight
=
Tools can assist counting input and output
weight loss/gain
Tools: Measuring Energy Output
Odometers measure
time, distance, velocity
Commonly used
in exercise
Provide rough estimate
of calories burned
Tools: Pedometer
A pedometer is worn
on the waist
Can be worn all day,
used anywhere
Measures steps, e.g.
during exercise
Tools: Measuring Energy Intake
Manually counting calories
Pre-packaged foods
and serving sizes
Calorie or food diary
None of these are easy to use consistently over long periods of time
Our concept: Bite counter
Can vibrate to
tell you when to
stop eating
Worn like a watch
Automatically tracks how many
bites of food have been taken
How do bites relate to calories?
Weight loss does not happen in a single meal (or bite); it takes
weeks to months of consistently consuming less food
1 pound per week is a common guideline (American Heart Association)
People are bad at counting calories
Undercounting 200-500 calories per day (Champagne et. al. 2002)
Calories are not as accurate as we like to believe
Frozen dinners 8% under, fast food 18% under (Urban et. al. 2010)
People tend to eat the same foods week-to-week
Reduce the bites … reduce the calories
Does bite size matter?
How many bites do you eat in a meal?
10-75, based upon our testing
Bite count goals will be customized to the individual, and will be
based upon monitoring the individual for a week
Example: After a week of monitoring, we determine your average
lunch bite count is 38; we recommend reducing that to 32
At daily and weekly intervals, goals could be given to smooth out
spikes in eating behavior
Example: Upon reaching dinner, the device indicates you have
already consumed 183 bites today; a small dinner is recommended
Advantages of our device
Objective
No guessing, or thinking back over a day to total
consumption
Automated
You can be doing other things (talking with friends,
working, watching TV, etc.); the device does all
the counting
Real-time feedback
The device can give you cues to stop eating
BEFORE you have consumed more than you
intended
How it works
The wrist undergoes a characteristic roll motion
during the taking of a bite of food
Biologically, this can be related to the necessary orientations for
(1) picking food up, and (2) placing food into the mouth
How it works
Tracking the wrist roll motion, we have identified a 3-event
pattern that corresponds to the taking of a bite of food
Demo
How well does it work?
Experiment #1: Waffles
51 subjects, 139 meals, same food (waffles), same utensil
94% bites correctly detected
80% positive predictive value
Experiment #2: Any foods, in lab
47 subjects, 49 meals, any food (and drink), any utensil
88% bites correctly detected
76% positive predictive value
Talking, other actions during 67% of bites
Experiment #3: Two weeks
Test relationship between bites and calories
•
•
•
•
83 subjects wore for 2 weeks
3246 meals
76% have correlation > 0.4
69% between 0.4 and 0.7
Correlation: 0.4
Correlation: 0.7
3/21/2016
17
Experiment #4: Cafeteria
Harcombe Dining Hall; seats 800 people
3/21/2016
Instrumented dining table
18
How do we build it?
Embedded System Design
Lab model
Watch model
Breadboarding and parts selection
I/O (buttons, speaker, vibratory
motor, USB port)
gyroscope
LCD
charger
processor
battery, holder
Processor
development board
dev board JTAG-USB
LCD
surfboard and custom breakout board
Gyroscope
typically follow “application circuit” from manual
Battery
What can you wear?
How long will it run?
Design stages
Breadboard
Circuit diagrams
Design stages
Circuit diagrams
Printed circuit board (PCB) & manufacturing
EEEA, Inc
Design stages
PCB
Molding and case production
Center Line
Technology
Design stages
PCB, case
Assembly
Bite counter PCB
speaker
gyroscope
LCD
charger
USB port
battery
processor
JTAG
buttons
USB-RS232 bridge