在這個快速變化的世界中保持自我健康很重要。跑步機對於那些試圖保持健康和穩定的人有明顯的好處。在運動期間分析自我健康很重要。使用人工智能（AI）或機器智能作為跑步機上跑步者的私人健身教練是在運動期間自我可靠的更好方法。非接觸式熱成像或熱成像使用人體的紅外圖像來獲得對角線指示。熱成像相機或紅外相機使用IR輻射並形成熱區圖像。使用紅外熱像儀是一種快速，無創的人體溫度監測方法。因此有助於實時分析跑步機上跑步者的健康狀況。
該研究的重點是為在跑步機上跑步的人構建基於卷積神經網絡（CNN）的個人物理訓練器。 在跑步者開始跑步之前，系統會向跑步者詢問年齡，性別等信息。 該系統不斷通知每個階段的運動，在緊急情況下提供指導，正確的指導和警報。 研究還集中在跑步者的生理狀態，比較跑步機上跑步的效果和體溫，心跳，卡路里燃燒和肌肉拉傷等不同特徵。

To maintain self-healthy in this fast moving world is important. Treadmills have a distinct benefit for the ones trying to keep themselves healthy and steady. It is important to analyze self-health during exercise. Using Artificial Intelligence (AI) or machine intelligence as a personal fitness trainer for runners on treadmill is a better way to be self-reliable during exercise. Non-contact thermography or thermographic imaging uses infrared images of the human body to get diagonostic indications. A thermographic camera or infrared camera uses IR radiation and forms a heat zone image. Using Infrared thermal camera is a fast and non-invasive way to monitor human body temperature. Thus helps in real time analysis of health of a runner on treadmill.

This research focuses constructing a Convolutional Neural Network (CNN) based personal physical trainer for individuals running on a treadmill. Before a runner starts running, the system will ask the runner for some information like age, gender. This system keeps notifying each stage of exercise, giving guidelines, proper instruction and alerts at the time of emergency. The research also focusses on the physiological status of runners comparing the effect of running on a treadmill and different features like body temperature, heartbeat, calories burnt and strain on muscles.

Body temperature increases with increase in intensity of exercise and leads to dehydration. Temperature continues to rise above 38ºC while running. The temperature keeps rising when an individual keeps running without fluid intake and may leed to dizziness. To avoid such situation, the system gives an alert for fluid intake, to reduce the intensity of exercise and take rest at situations that are more critical (above 39.4ºC, causing fever). In terms of heart rate, the experiment results show that heart rate inreases while running.

To provide more oxygen and nutrients, the blood circulates in the muscles. Heart rate increases 10% without fluid intake and only 5% with fluid intake. Heart rate increases to maximum heart rate (MHR) and is reserved. MHR is different for different individuals. This sytem also provides an indication for an individual’s MHR. In addition, the results showed that the heart rate return soon to normal for people who are healthy. Calories burnt increased with the duration of exercise. Vigourous exercise intensity results in strain on muscles. The system alerts when there is stress on a muscle.

Chinese Abstract........................................................i
English Abstract........................................................ii
Aknowledgment........................................................iii
Table of contents........................................................iv
List of Tables........................................................vii
List of figures........................................................viii
Chapter I. Introduction...................................................1
1.1 Research motivation..........................................1
1.2 Research purposes..........................................1
Chapter II. Literature Review..........................................2
2.1 Treadmill..........................................2
2.1.1 Treadmill History..........................................2
2.1.2 Different Treadmills..........................................2
2.1.2.1 Ancient Roman Tread Mill Crane..........................................2
2.1.2.2 The Horse Treadmill..........................................2
2.1.2.3 Dog, Sheep and Goat Treadmills..........................................3
2.1.2.4 Medical Treadmills..........................................3
2.1.2.5 Treadmills for Exercise..........................................3
2.1.2.6 Life Fitness..........................................3
2.2 Artificial Intelligence (AI)..........................................4
2.2.1 History of Artificial Intelligence..........................................4
2.2.2 Machine learning..........................................4
2.2.2.1 Supervised learning..........................................5
2.2.2.2 Unsupervised learning..........................................6
2.2.3 Deep Learning..........................................7
2.2.4 Convolutional neural network (CNN)..........................................8
2.2.5 Types of Layers in CNN..........................................9
2.2.5.1 Convolutional Layer..........................................9
2.2.5.2 Pooling Layer..........................................10
2.2.5.3 Fully Connected Layer..........................................11
2.3 Infrared Thermography..........................................12
2.3.1 Discovery of Infrared..........................................12
2.3.2 Thermographic camera..........................................13
2.3.3 Types of thermographic cameras..........................................13
2.3.3.1 Cooled infrared detector..........................................13
2.3.3.2 Uncooled infrared detectors..........................................13
2.3.4 Working principle of IR camera..........................................14
2.3.5 Applications of IR camera..........................................15
2.4 Human phisiology and running..........................................16
2.4.1 Body temperature..........................................16
2.4.1.1 Hypothalamus..........................................17
2.4.2 Heart rate..........................................18
2.4.3 Exercise and muscles..........................................19
Chapter III. Materials and methodology..........................................21
3.1 Hardwares used..........................................21
3.1.1 MLX90640 Themal camera sensor..........................................21
3.1.2 Heart rate sensor..........................................22
3.1.3 Heart rate based calorie expenditure calculator..........................................24
3.1.4 EMG muscle sensor..........................................25
3.1.5 Arduino Uno board..........................................28
3.1.6 Teensy 3.6..........................................29
3.1.7 TCA9548A multiplexer..........................................30
3.1.8 Raspberry Pi..........................................32
3.2 Softwares used..........................................33
3.2.1 Arduino micro-controller software..........................................33
3.2.2 Teensyduino..........................................33
3.2.3 Tensorflow..........................................34
3.2.3.1 Keras..........................................35
3.2.4 Python..........................................37
3.2.5 Processing..........................................38
3.3 Construction of convolutional neural network..................................41
3.3.1 Preprocessing..........................................41
3.3.2 Implementation..........................................41
3.4 Multiple MLX90640 cameras |(2x2 array) for better images......................46
3.4.1 Changing I2C address..........................................47
Chapter IV. Results and discussion..........................................49
4.1 CNN for analyzing IR camera images..........................................49
4.1.1 Accuracy & loss test for CNN..........................................50
4.1.1.1 CNN with 25 epochs..........................................50
4.1.1.2 CNN with 5 epochs..........................................51
4.1.2 Prediction of test samples by CNN..........................................52
4.1.2.1 Prediction of test samples with normal body temp..........................................52
4.1.2.2 Prediction of test samples with too low temp..........................................53
4.1.2.3 Prediction of test samples with temp falling low..........................................54
4.1.2.4 Prediction of test samples with temp above normal & below high temperature................................55
4.1.2.5 Prediction of test samples with temp too high............56
4.1.2.3 Prediction of test samples with 2x2 array..................................57
4.1.2.3 Prediction of test samples (full body) with temp 2x2 array................58
4.2 Exercise intensity v/s body temp with/without fluid intake..................59
4.3 Body temp v/s heart rate with/without fluid intake..................60
4.4 Body temp v/s heart rate with same intensity on treadmill..............61
4.5 Heart rate for calories burnt.................................62
Chapter V. Conclusion........................................................63
References................................................64
Appendices........................................................66
Extended Abstract......................................................77

 

[1] Yoshua Bengio, 2009, “Learning deep architectures for AI” Foundations and Trends in Machine Learning, Foundations & Trends in Machine Learning Volume 2 Issue 1, 2-8.

[2] Tianmei Guo; Jiwen Dong; Henjian Li; Yunxing Gao, 2017, “Simple convolutional neural network on image classification”, 2017 IEEE, 721-724.

[3] Abbas K Abbas, Katrin Jergus, Konrad Heimann, Thorsten Orlikowsky, 2011, “Neonatal non-contact respiratory monitoring based on real-time infrared thermography”, BioMed Central, 3-8.

[4] Juana M. Morillas-Ruiz and Pilar Hernández-Sánchez, 2015, “Oxidative Stress and Antioxidant Defenses Induced by Physical Exercise”, Intech, 222-234.

[5] Eduardo Borba Neves, Raphael Martins Cunha, Claudio Rosa, Natacha Sousa Antunes, Ivo Miguel Vieira Felisberto, José Vilaça-Alves, Victor Machado Reis, 2016, “Correlation between skin temperature and heart rate during exercise and recovery, and the influence of body position in these variables in untrained women”, Elsevier ,70–76.

[6] Deuk-Ja Oh, Hyeon-Ok Hong, and Bo-Ae Lee, 2016, “The effects of strenuous exercises on resting heart rate, blood pressure, and maximal oxygen uptake”, Journal of exercise rehabilation, 42-45.

[7] Seol-Jung Kang, Eon-ho Kim, PhD,2 and Kwang-Jun Ko, 2016, “Effects of aerobic exercise on the resting heart rate, physical fitness, and arterial stiffness of female patients with metabolic syndrome”, Journal of physical therapy science, 1764-1767.

[8] Cinzia Perrino, Giuseppe Gargiulo, Gianluigi Pironti , 2011, “Cardiovascular effects of treadmill exercise in physiological and pathological preclinical settings”, Am J Physiol Heart Circ Physiol, 1983–1989.

[9] Michael J. Joyner and Darren P. Casey, 2015, “Regulation of Increased Blood Flow (Hyperemia) to Muscles During Exercise: A Hierarchy of Competing Physiological Needs”, physiol Rev, 549-601.

[10] Sakellariou GK, Jackson MJ, Vasilaki A, “Redefining the major contributors to super‐ oxide production in contracting skeletal muscle.”, Free Radic Res., 12-29.

[11] U. Proske, D. L. Morgan, 2001, “Muscle damage from eccentric exercise: mechanism, mechanical signs, adaptation and clinical applications”, Journal of Phisiology, 333-345.

[12] Sandeep Chandra Bollepalli, Sastry S Challa, Soumya Jana, 2018, “Robust Heartbeat Detection from Multimodal Data via CNN-based Generalizable Information Fusion”, IEEE Trans Biomed Eng.,

[13] Leslie H. Willis, Cris A. Slentz, Lori A. Bateman, A. Tamlyn Shields, Lucy W. Piner, Connie W. Bales, Joseph A. Houmard, William E. Kraus, 2012, “Effects of aerobic and/or resistance training on body mass and fat mass in overweight or obese adults”, J Appl Physiol, 1831–1837.

[14] G. Karthik Reddy, K. Lokesh Achari ,2015, “A Non Invasive Method for Calculating Calories Burned during Exercise using Heartbeat”, 2015 IEEE, 2-5.

[15] Andrea Guidi, Alberto Greco, Gaetano Valenza, Enzo Pasquale Scilingo, 2017, “Heart Rate Variability Analysis during Muscle Fatigue due to prolonged Isometric Contraction”, EEE Eng Med Biol Soc., 1324-1327.

[16] Alan G smith, 2011, “Introduction to Arduino”, Introtoarduinoebooks, 59-73.

[17] Michael McRoberts, 2007, “Beginning Arduino second edition”, ITebooks, 3-12.

[18] Guif ang Lin, Wei Shen, 2018, “Research on convolutional neural network based on improved Relu piecewise activation function”, Elsevier, 978-984.

[19] Tensorflow tutorial, 2018, “Image classifier using convolutional neural network” Tensorflow CV tricks, 1-4.

[20] Volodymyr Mnih1, Koray Kavukcuoglu1, David Silve, 2018, “Human-level control through deep reinforcement learning”, Research Lettter, 529-533.

[21] Shyam P. Adhikari1, Changju Yang, Hyongsuk Kim, 2016 “Construction of CNN Template via Learning with Random Weight Change Algorithm”, CNNA, 21-22.

[22] Martin Thoma, 2017, “Analysis and Optimization of Convolutional Neural Network Architectures”, Master thesis, 3-10.

[23] Diederik P. Kingma, Jimmy Lei Ba, 2015, Adam, A method for stochastic optimization”, ICLR, 1,2,7.

[24] Adeel Javed, 2016, “Building Arduino projects for the internet of things”, ISBN-13(electronic), 15-26.

[25] Markus Jangblad, 2018, “Object Detection in Infrared Images using Deep Convolutional Neural Networks”, Uppsala UPTEC, 34,35,36, 37,39.