Fitomatic: A Web Based Automated Healthcare Supervision and Monitoring App

Fitomatic: A Web Based Automated Healthcare Supervision and Monitoring App

1Student, Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai, India.

2Student, Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai, India.

3Student, Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai, India.

4Student, Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai, India. Assistant Professor, Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai, India. ***

Abstract - In recent times, people all around the world have realized the importance of maintaining a healthy lifestyle. Thus, more and more people have decided to focus on having a healthy and fit body. However, monitoring health and fitness without proper assistance could be confusing and difficult. To achieve a healthy and fit body, not only exercise but also a proper diet should be followed. Also, every individual has different fitness goals and thus, for every individual the fitness regimen differs. Keeping all these constraints in mind, we have proposed a system that helps beginners as well as fitness enthusiasts take the first step to achieve their fitness goals. Our solution aims to help individuals improve their quality of life, by recommending healthier diet and exercise plans by analyzing their BMI and monitoring the exercises done by the user. Since many individuals cannot make time out of their busy schedules to visit the gym, this system is beneficial to them because they can perform exercises and also get them monitored virtually without the need of a physical trainer.

Key Words: Machine Learning, MediaPipe, KNN, Fitness, Pose Detection, Recommendation, BMI.

1.INTRODUCTION

In our work, we introduce Fitomatic, a web app which tracks the fitness activities of users, their diet and meal tracking,and detects theusers exerciseposture.Owing to busy schedules and work pressure people are not paying attention to their health and fitness. Physical inactiveness is the most important problem in today’s generation. It is importanttounderstandthatdietandexercisevariesfrom users having different lifestyles, height, weight, sex, age, and activity level, however diet and exercise are both correlated.

1.1 Importance of Fitness

The importance of having good physical fitness cannot be stressed any further in the times that we find ourselves right now. People have been struggling with various

health-related problems [1] such as eye strain, mental stress, irregular sleep patterns, obesity, decreased immunity, etc. Immense emphasis has been put on by bodies like WHO (World Health Organization) since the spread of COVID-19 started increasing, on improving our health and immunity for being safe from the coronavirus andproperdietandexerciseplaysapivotalroleinmaking our bodies healthier. Some mobile applications provide expertsupport andsessions ona paidbasistogeta more personalizedandfocusedoptionfortrainingandguidance. Thus,aproductthatisfreeofcostisneededsothatitcan beusedbyall.

1.2 Research Studies

Although people are becoming more and more healthconscious, they still do not have the time to dedicate to going to the gym. This explains why working people all around the world prefer health and fitness tracking apps. Recent Statistical studies show that within the first week of lockdown, the Daily Active Users (DAU) in Health & FitnessAppscategorysawanupsurgeofalmost14%.This led to a tremendously high download growth rate, nearly 157%wasobservedin-homefitnessappsinIndia[2].

Therefore, a method is required which is much more accessible and at the same time, is reliable. In this work, weaimto:

● Providea platformtosatisfyall ofusers’needsat oneplace.

● Provide accurate and proper training, all at the convenienceofusers.

● Provide constant feedback to improve the quality ofperformanceofusers.

● Provide healthy diet plans which suit the user, taking into consideration their allergies and workoutregimen.

2. RELATED WORK

A decent amount of work has been done for developing designsforhealthmonitoringapplications.In[3],asystem is proposed which can help doctors to recommend diet andexercisetothepatients. Dealswithhealthmonitoring of disease like diabetes etc. based on patients’ latest reports using the Machine learning Technique i.e C4.5. They conclude C4.5 is better than the ID3 algorithm with respecttoboththedata-setsthatwereused.

S. Agarwal et al. [4], designed an application called FitMe, which aims to reduce the dependency on actual trainers and provide health benefits anywhere, anytime, free of cost and with limited hardware support. FitMe utilizes lightweight deep learning models for accurate pose estimation of the users. In addition to checking the accuracyofposes,itprovidesinstantfeedbacktousersso that they can maintain the right postures on the fly. The quality results obtained are shown in this work and further proved that it has massive scope for adoption by peoplefortheirfitnessneedsbeinginsidetheirhomes.

In [5] Gourangi Taware, Rohit Agrawal, Pratik Dhende , Prathamesh Jondhalekar, Shailesh Hule, introduce Fitcercise,anapplicationthatdetectstheexerciseposition oftheuser counts the prescribed exercise repetitionsand gives individualized, comprehensive analysis about enhancingtheuser'sbodyposture.

D. Shah, V. Rautela, C. Sharma and A. Florence A, "Yoga Pose Detection Using Posenet and k-NN [6] designed a project that carries a non-profit system that strives to develop core muscles using yoga-like poses. Virtual yoga asana practice is possible thanks to the totally accurate position detection provided by the proposed method. The cosine similarity technique is used to consider the deviation of the angle created with the original values. This study uses computer vision algorithms and the open pose to evaluate human poses and a person's yoga stance (open-source library). The proposed model was trained with 90% of data and tested with 10% of the same with real-timetesting,resultingin94%accuracy.

A. Singh, S. Agarwal, P. Nagrath, A. Saxena and N. Thakur [7], an article that covers the problems with estimating human posture and provides an overview of extensive research on the subject, including deep learning methodology and conventional image-based algorithms, hasbeenoffered.Theauthorhascreatedastraightforward model using a convolutional neural network that estimates the postures and exemplifies the potential of CNNs after examining numerous findings and identifying theconstraints.

An application is designed by Prof. Prajkta Khaire, RishikeshSuvarna,AshrafChaudharyin[8],thatprovides

the user with a complex algorithm which can provide the user witha dietplan based onhis/hercharacteristicslike height, weight, BMI. With just one button click, users will be able to register an account, manage their account, and access the diet through the suggested application's userfriendly User-Interface. It also offers the option to get in touch with a real nutritionist for advice if the user has a foodallergy.

In another work presented by A. Henning, B. Alvarez, C. Brady, J. Kopec and E. Tkacz [9], have designed a ElastoTrak that combines the cardiovascular workout of a treadmill with the resistance training of springs, thereby enabling users to achieve the benefits of both exercises simultaneously. The strength of the device's frame, the device's ability to successfully boost the user's heart rate into the cardiovascular training range, and the device's usabilitywillallbetested.

Usinga professional workoutasa reference,Nagarkoti, R. Teotia, A. K. Mahale, and P. K. Das suggested a system in [10] to analyze a user's body position during exercise. In order to identify mistakes and offer the user corrective action, we depict the human body as a collection of limbs andexamineanglesbetweenlimbpairs.

Last but not least, S. Bian, V. F. Rey, P. Hevesi, and P. Lukowiczstudiedthepotentialofthissensorymodalityin gym workouts in [11], where they also detailed the physical theoryunderlyingthepervasiveelectriccoupling betweenthehumanbodyandsurroundings.

2.1 Limitations

● Some mobile applications provide expert support and sessions on a paid basis to get a more personalizedandfocussedoptionfortrainingand guidance.

● Tedious task of searching for integrity in the manualsystemsbefore.

● All existing systems are not well integrated. Rathertheyaregoodintheirownrespectedwork.

● ExistingappsthatusedMLmodelsformonitoring would only be able to estimate or identify the posefromastaticimage.

● Generation of the feedback in the form of paragraphs.

● Complex hardware infrastructure is neither affordablebyusers,noriseasytouse.

2.2 Problem Statement

Mostusersmustutilize various applicationstokeep track of their workouts, routines, and diet preparation. Consumerseventuallyloseinterestsincetheyfinditquite difficult to use several apps and maintain track of it. Although people are becoming more and more healthconscious, they still do not have the time to dedicate to going to the gym. This explains why working people all aroundtheworldpreferhealthandfitnesstrackingapps.

3. PROPOSED SYSTEM

3.1. System Design

Toachievethedesiredgoalofrecommendingpersonalized diets along with exercise tracking, we use the following methodology, containing two phases; Phase 1: Diet recommendation and calorie tracking, Phase 2 : Exercise livemonitoringandfeedbackgeneration.

suggestionsonanitem'squalitiesorcontentisknownasa content-based recommendation engine. It works by analyzing the content of items, such as text, images, or audio, and identifying patterns or features that are associated with certain items. The following step involves comparinggoodsandsuggestingcomparableonestousers usingthesepatternsorattributes.

Theprocedureisasfollows:

a) Taking user Data: Starting with entering patient’s detailssuchasheight,weight,age,gender,activitylevel.

b) Calculating BMI: Calculation of BMI and calories required with formula using the personal details taken as input.

BMI (Body Mass Index) and Calories Requirement Calculation

BMI = [Body Weight (Kg)]/[Sq of body weight in m] =kg/m^2Where,Underweight<18.5

NormalWeight=18.5-24.9

Overweight=25-29.9

Obesity>30Calories:

ForMen:66.5+13.8(W)+5.0(H)-6.8(A)

ForWomen:66.51+9.6(W)+1.9(H)-4.7(A)

Where,W=Weightinlbs.H=Heightininches.A=Agein years.

c) Content – based Filtering: The Recommendation engine uses information about the nutritional values and ingredients of foods to make personalized recommendations to users. Also, it takes into consideration an individual's dietary restrictions and preferences,suchasallergiesorfoodpreferences.

d) Recommendation: Users are provided with a customized exclusive experience which will help them make better choices about what to eat and improve their overallhealththatisadietisrecommended.

:SystemArchitecture

3.1.1. Diet Recommendation and Calorie Tracking

Diet recommendation is implemented using a contentbased approach. A recommendation engine that bases its

Fig -2:DietRecommendationandCalorieTracking

3.1.2. Exercise Recommendations and Pose Detection

Monitoringofuserexercisesisdonebythemethodofpose estimation. Pose estimation refers to a computer vision techniquethatdetectsandtrackshumanfiguresorobjects in videos and images. In the case of humans, it could help determinethelocationofthekeybodypoints.

Fig -3:ExerciseMonitoringandPoseDetection

3.2 Framework/Algorithm

3.2.1. Nearest Neighbor for Recommendation

The Nearest Neighbors model is utilized in the diet recommendation section for prediction, with the cosine metric beingusedforcategorical data andthe brute force techniquebeingemployedforathoroughsearch.TheKNN model will curate a diet in accordance with the nutrient limitreceivedfromtheuserandadviseit.

Basedontheirnutritionalvalue,locatethe foodsormeals thataretheclosesttoaspecificfoodormeal.

Nearest neighbors can be used in a diet recommendation systemtodeterminewhichfoodsarethemostcomparable intermsofnutrients.Theconceptisthatiftwofoodshave comparable nutrient profiles for example, comparable levels of protein, fat, carbs, vitamins, and minerals then they are probably going to have comparable impacts on thebodyintermsofnutritionandhealth.

In our project, we use a pre-trained KNeighborsClassifier on the data to unsupervised identify the samples that are mostcomparable.

The fig explains the deviation and distribution of the data pointsfromanormaldistributionandaccordingtothetest input, most similar samples from the dataset are recommended.

3.2.2. Mediapipe holistic Framework

Mediapipe Holistic Framework enables live perception of simultaneous human pose, face landmarks, and hand tracking in real-time. It integrates separate models for pose, face and hand components, each of which are optimizedfortheirparticulardomain.Itisknowntooffer fastandaccurate,yetseparate,solutionsforthesetasks.

Thestepstoidentifyasuccessmovementare:

a) Phone camera to capture a (or a series of) real-time images.

b) The python module then identifies the users’ skeleton andjointpositionfromthecapturedimages.

c) When the skeleton and joint positions are pinned, the successofamovementiscalculated.

d) Ifthemovementisasuccess,thenumbercounts.Once a set of work-out is done, the record is refreshed and kept for further advice. The fitness records that show one’simprovementandachievementscanbeusedfor furtheradvice.

4. CONCLUSIONS

This work proposes an application designed specifically forfitnessenthusiasts.Byutilizingawebcamera,machine learning modules and recommendation engines can help usersachievetheirfitnessgoalsallatoneplace.Thefuture work would consist of a system for tracking of diet and exercise and in continuation would provide alternate options with respect to the user’s ailments to a particular fooditemorexerciseincaseofchangeofuserpreferences and creating a regular and emergency alert system to remind the user before every follow-up session and in alertuserincasesofextremereports.

ACKNOWLEDGEMENT

We wish to state that the work embodied in this project titled “Fitomatic: A Web Based Automated Healthcare Supervision and Monitoring App” forms our own contributiontotheworkcarriedoutundertheguidanceof 'Prof. Dilip Dalgade’s direction at MCT's Rajiv Gandhi Institute of Technology. We affirm that this written submissioncontainsourideasinourownwords,andthat whenotherpeople'sthoughtsorwordsareused,theyare properlyacknowledgedandcited.

REFERENCES

[1]"Covid-19lockdownhasnegativelyimpactedkids’diet, sleep and physical activity: Study", The Indian Express, 2020. [Online].Available: Covid-19 lockdown has negatively impacted kids’ diet, sleep and physical activity:Study|LifestyleNews,TheIndianExpress

[2]C.Ang,"Fitnessappsgrewbynearly50%during the firsthalf of2020,studyfinds",WorldEconomicForum, 2020.[Online].Available: Fitness app downloads grew by 46% worldwide in COVID-19 | World Economic Forum(weforum.org)

[3] D. Mogaveera, V. Mathur and S. Waghela, "e-Health Monitoring System with Diet and Fitness Recommendation using Machine Learning", 2021 6th International Conference on Inventive Computation Technologies(ICICT),2021.

[4] S.Agarwal et al., "FitMe: A Fitness Application for Accurate Pose Estimation Using Deep Learning", 2021 2nd International Conference on Secure Cyber ComputingandCommunications(ICSCC),2021.

[5] Gourangi Taware , Rohit Agrawal , Pratik Dhende , PrathameshJondhalekar,ShaileshHule,2021,AI-based WorkoutAssistantandFitnessguide,INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY(IJERT)Volume10,Issue11(November 2021)

[6]D.Shah,V.Rautela,C.SharmaandA.FlorenceA,"Yoga Pose Detection Using Posenet and k-NN," 2021 International Conference on Computing, CommunicationandGreenEngineering(CCGE),2021

[7] A. Singh, S. Agarwal, P. Nagrath, A. Saxena and N. Thakur, "Human Pose Estimation Using Convolutional Neural Networks," 2019 Amity International ConferenceonArtificialIntelligence(AICAI),2019

[8] Prof. Prajkta Khaire, Rishikesh Suvarna, Ashraf Chaudhary, “Virtual Dietitian: An Android based Application to Provide Diet”, International Research Journal of Engineering and Technology (IRJET), Volume:07Issue:01|Jan2020

[9]A.Henning,B.Alvarez,C.Brady,J.Kopec andE.Tkacz, "Workout Machine that Combines Cardiovascular Exercise with Strength Training," 2020 39th Annual NortheastBioengineeringConference,2020

[10] A. Nagarkoti, R. Teotia, A. K. Mahale and P. K. Das, "Realtime Indoor Workout Analysis Using Machine Learning & Computer Vision," 2019 41st Annual International Conference of the IEEE Engineering in MedicineandBiologySociety(EMBC),2019.

[11] S.Bian,V.F.Rey,P.HevesiandP.Lukowicz,"Passive CapacitivebasedApproachforFullBodyGymWorkout Recognition and Counting," 2019 IEEE International Conference on Pervasive Computing and Communications.

[12]B.Sainz-De-Abajo,J.M.García-Alonso,J.J. BerrocalOlmeda, S. Laso-Mangas and I. De La Torre-Díez, "FoodScan: Food Monitoring App by Scanning the GroceriesReceipts.”

BIOGRAPHIES

KingshukDebnath,Undergraduate Student, BE Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai University, Mumbai.

kingshuk.d16@gmail.com

AnushaSunilkumar,Undergraduate Student, BE Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai University, Mumbai.

anushasunil71201@gmail.com

NehaBhange,Undergraduate Student, BE Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai University, Mumbai. neha05bhange@gmail.com

ElrishaDsilva,Undergraduate Student, BE Computer Engineering, MCT’s Rajiv Gandhi Institute of Technology, Mumbai University, Mumbai. elrishad30@gmail.com

Dilip Dalgade, Assistant Professor, Computer Engineering, Expertise in Data Structures and Algorithm and Machine Learning MCT’s Rajiv Gandhi Institute of Technology, Mumbai University,Mumbai. dilip.dalgade@mctrgit.ac.in