smart-log: an automated, predictive nutrition monitoring ... · overview of iot based smart-log...

of 27 /27
Smart - Log: An Automated, Predictive Nutrition Monitoring System for Infants through IoT P. Sundaravadivel 1 , K. Kesavan 2 , L. Kesavan 3 , S. P. Mohanty 4 , E. Kougianos 5 , and M. K. Ganapathiraju 6 University of North Texas, Denton, TX 76203, USA. 1,4,5 University of Texas, Arlington, TX , USA. 2 Independent Researcher, Dallas, TX, USA. 3 University of Pittsburgh, Pittsburgh, PA, USA. 6 Email: [email protected] 1 , [email protected] 2 , [email protected] 3 , [email protected] 4 , [email protected] 5 , [email protected] 6 12th Jan 2018 Smart-Log Talk, ICCE 2018 1

Author: others

Post on 22-Aug-2020

1 views

Category:

Documents


0 download

Embed Size (px)

TRANSCRIPT

  • Smart-Log: An Automated, Predictive Nutrition Monitoring System for Infants

    through IoT

    P. Sundaravadivel1, K. Kesavan2, L. Kesavan3, S. P. Mohanty4,

    E. Kougianos5, and M. K. Ganapathiraju6

    University of North Texas, Denton, TX 76203, USA.1,4,5

    University of Texas, Arlington, TX , USA.2

    Independent Researcher, Dallas, TX, USA.3

    University of Pittsburgh, Pittsburgh, PA, USA.6

    Email: [email protected], [email protected],

    [email protected]il.com3, [email protected],

    [email protected], [email protected]

    12th Jan 2018 Smart-Log Talk, ICCE 2018 1

  • 2

    Introduction

    Novel Contributions

    System level Modeling of Smart-Log System

    Implementation and Validation

    Conclusions and Future Research

    Outline of the talk

    12th Jan 2018 Smart-Log Talk, ICCE 2018

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 3

    Introduction

    Nutrition imbalances in infants

    Significance of Nutrition Monitoring System

    IoT in Smart Healthcare

  • Nutrition Imbalances

    • Imbalances in nutrition can occur either due to

    undernourishment or over nourishment.

    • Malnourished children tend to have weakened

    immune system.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 4

  • Significance of Nutrition Monitoring System

    • Healthy lifestyle starts from having a well balanced

    diet.

    • Many children tend to be fussy eaters where

    aversion of certain food stays forever.

    • In such scenarios, finding potential replacement of

    the concerned food item is important to maintain a

    well balanced diet.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 5

  • Internet of Things (IoT)

    • The Internet of Things helps in connecting real

    world sensor data to cloud based solutions.

    • The IoT acts as a virtual brain

    of wireless sensor networks

    which can be realized as mixed

    signal systems.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 6

  • Novel Contributions

    • An automated solution which keeps track of the

    amount of food consumed along with the nutrition

    facts.

    • A prediction based method to analyze nutrient

    values of future meals.

    • Feedback is taken from the user each time to

    customize the user’s need.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 7

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 8

    System Level Modeling of Smart-Log System

    Smart Sensor Board

    Data Acquisition

    Future Meal Predictions

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 9

    Overview of IoT based Smart-Log

    System

    Box-1 Box-2 Box-3

    Box-4 Box-5 Box-6

    Box-7 Box-8 Box-9

    Piezo-sensor

    Food

    Product

    Data logged into the cloud

    Feedback

    to the user

    Camera to acquire

    nutrient values

  • Smart Sensor board

    • The proposed sensor system consists of

    piezoelectric sensors paired with a micro-controller.

    • Piezoelectric sensors generate equivalent voltage

    signals for the applied weight or mechanical force.

    • This equivalent voltage value can be read with help

    of micro-controller which saves the obtained value

    along with the timestamp.

    System level Modelling of Smart-Log System

    12th Jan 2018 Smart-Log Talk, ICCE 2018 10

  • System-Level Modelling of Smart-Log System

    Smart Sensor board

    • Micro-controller helps in scheduling

    • It helps in easier integration into wireless modules,

    which helps in connecting the prototypes to the

    internet for IoT based applications.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 11

  • Data Acquisition in Smart Log

    System

    Optical Character Recognition

    Application Program Interface (API)

    Database Approach

    12th Jan 2018 Smart-Log Talk, ICCE 2018 12

  • User takes a picture of the Nutrition

    Facts using Smart Phone

    Use Optical Character Recognition

    (OCR) to convert images to text

    Nutrition facts obtained through

    OCR

    User scans the barcode of the product

    Using Open Application Program

    Interface (API)’s and Database

    approach, the nutrition facts are

    acquired from Central database

    Nutrient facts obtained through

    API’s

    Weight and Time information obtained

    through Sensing Board

    Calculate Nutrient Value of the meal

    Save the Nutrient value, Weight, Time

    of each meal for future predictions

    Data Acquisition in Smart-Log System

    12th Jan 2018 Smart-Log Talk, ICCE 2018 13

  • System level Modelling of Smart-Log System

    Future Meal Predictions

    • The main concern in infants is the amount of food

    wasted after each meal.

    • The purpose of each meal may vary from person

    to person.

    • By considering the weight of the wasted food, the

    corresponding nutrient value is calculated.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 14

  • Calculate weight of the left over food

    Calculate the left over nutrient value

    Ask the User regarding Purpose

    of the meal

    1. Rich in Carbohydrates

    2. Rich in Protein

    3. Rich in Fiber

    4. Well-balanced Diet

    Goal Achieved ?

    Give feedback and Predictions for the

    next meal

    Calculated

    Nutrient values for

    the meal

    Predictions for the Next Meal

    12th Jan 2018 Smart-Log Talk, ICCE 2018 15

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 16

    Implementation & Validation

    Sensor Design

    Classifier evaluation using WEKA

  • • The sensor system was implemented using Proteus

    and the data analysis to build an efficient

    prediction based system is done using WEKA.

    • With help of a JAVA program, the input and output

    of the classifier was displayed in a webpage.

    Implementation of Smart-Log System

    12th Jan 2018 Smart-Log Talk, ICCE 2018 17

  • The sensor system is required to possess the

    following functionalities:

    • Calculate the total weight of the product placed on

    the sensor board.

    • Calculate weight of the product after the user’s

    consumption.

    • Log the weight each time in order to give better

    predictions.

    Implementation of Smart-Log System

    12th Jan 2018 Smart-Log Talk, ICCE 2018 18

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 19

    ARDUINO based Sensor Module

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 20

    ARDUINO based Sensor Module

    • ARDUINO board helps in faster prototyping and

    can be easily programmed using C and C++.

    • A LCD module was used in order to read the

    variations in voltage value in accordance to the

    weight of the elements.

    • Data logging was achieved through the USB port.

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 21

    Data for the Sensor System

    • A database of 8791 instances consisting of both

    readily available and raw ingredients for baby food

    was considered.

    • This SR8 database is available through the US

    Department of Agriculture website, and can also

    be accessed using API.

    • The unique NDB number is used to access the food

    item.

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 22

    • Waikato Environment for Knowledge Analysis

    (WEKA) helps in evaluating the system using

    numerous algorithms.

    • Whenever a user makes a data entry, an Attribute-

    Relation File Format (.arff) is created, which serves

    as input for WEKA.

    Classifier evaluation using WEKA

  • Classifier evaluation using WEKA

    Classifiers Mean

    Absolute

    error

    RMS error Relative

    absolute

    error (%)

    Root

    Relative

    Squared

    error (%)

    Naïve Bayes 0.0106 0.0179 2.65 3.98

    Bayes Net 0.0052 0.0055 1.3108 1.2236

    Naïve Bayes Updatable 0.0106 0.0179 2.65 3.98

    Multilayer Perceptron 0.0149 0.0179 3.72 3.99

    Decision Table 0.0322 0.0363 9.06 8.63

    Simple Logistic 0.1192 0.1192 33.55 22.28

    Naïve Bayes Multinomial Text 0.0354 0.0419 4.78 5.06

    12th Jan 2018 Smart-Log Talk, ICCE 2018 23

  • Performance Comparison of Food recognition System

    Research Works Food recognition method Efficiency

    (%)

    Chen et al. [23] Image based feature extraction 90.9

    Maiti et al. [19] Image based feature extraction 96.2

    Joutout et al [12] Image based classification

    using multiple kernel learning

    61.34

    This paper Mapping nutrition facts to a

    database

    98.4

    12th Jan 2018 Smart-Log Talk, ICCE 2018 24

  • Characterization Table for the Proposed

    System

    Characteristics

    Sensor System ARDUINO UNO, Piezoelectric sensor and LCD module

    Data acquisition OCR, API and Database approach

    Data Analysis Tool WEKA

    Input Dataset 8172 instances

    Classifier Naïve Bayes

    Accuracy (Worst case) 98.4%

    12th Jan 2018 Smart-Log Talk, ICCE 2018 25

  • Conclusion and Future Research

    • This paper presents an autonomous food data logging

    system.

    • Two approaches for data acquisition have been evaluated

    using WEKA in order to build a better prediction model.

    • As a cost effective sensor system, with seamless data

    logging method, this system can be an essential consumer

    electronic device used in child care.

    12th Jan 2018 Smart-Log Talk, ICCE 2018 26

  • 12th Jan 2018 Smart-Log Talk, ICCE 2018 27

    Thank You !!!Slides Available at: http://www.smohanty.org