(Đồ án hcmute) development of a wearable ai based device for wrist pulse diagnosis

MINISTRY OF EDUCATION AND TRAINING HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY AND EDUCATION FACULTY OF MECHANICAL ENGINEERING GRADUATION PROJECT ROBOTIC AND ARTIFICIAL INTELLIGENCE DEVELOPMENT OF A WEARABLE AI-BASED DEVICE FOR WRIST PULSE DIAGNOSIS n ADVISOR: PhD BUI HA DUC STUDENT: DAO THANH QUAN LE QUOC TUAN SKL010837 Ho Chi Minh city, July 2023 MINISTRY OF EDUCATION AND TRAINING HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY AND EDUCATION FACULTY OF MECHANICAL ENGINEERING n GRADUATION THESIS DEVELOPMENT OF A WEARABLE AI-BASED DEVICE FOR WRIST PULSE DIAGNOSIS Supervisor: Student: Student ID: Student: Student ID: Year Of Admission: BUI HA DUC, PhD DAO THANH QUAN 19134081 LE QUOC TUAN 19134091 2019 - 2023 Ho Chi Minh City, July 2023 HO CHI MINH CITY UNIVERSITY OF TECHNOLOGY AND EDUCATION FACULTY OF MECHANICAL ENGINEERING DEPARTMENT OF MECHATRONICS n GRADUATION THESIS DEVELOPMENT OF A WEARABLE AI-BASED DEVICE FOR WRIST PULSE DIAGNOSIS Supervisor: Student: Student ID: Class: Student: Student ID: Class: Year Of Admission: BUI HA DUC, PhD DAO THANH QUAN 19134081 19134 LE QUOC TUAN 19134091 19134 2019 - 2023 Ho Chi Minh City, July 2023 TRƯỜNG ĐẠI HỌC SƯ PHẠM KỸ THUẬT TP HCM KHOA CƠ KHÍ CHẾ TẠO MÁY CỘNG HỒ XÃ HỘI CHỦ NGHĨA VIỆT NAM Độc lập - Tự – Hạnh phúc NHIỆM VỤ ĐỒ ÁN TỐT NGHIỆP Học kỳ II/ năm học 2022-2023 Giảng viên hướng dẫn: TS Bùi Hà Đức…………….……… ……… Sinh viên thực hiện: Lê Quốc Tuấn………………MSSV: 19134091…Điện thoại 0935920245 Đào Thanh Quân……………MSSV: 19134081…Điện thoại 0962600379 Mã số đề tài: 22223DT95 – Tên đề tài: Development of a Wearable AI-based Device for Wrist Pulse Diagnosis Các số liệu, tài liệu ban đầu: …………….……… ……….…………………………………………………………… …………….……… ……….…………………………………………………………… n Nội dung đồ án: Xây dựng thiết bị đo phần mềm xử lý tín hiệu mạch đập cổ tay, qua chẩn đốn tình trạng sức khỏe người …………….……… ……….…………………………………………………………… Các sản phẩm dự kiến Thiết bị đo mạch đập cổ tay Phần mềm phân tích chẩn đốn tình trạng sức khỏe người dựa tín hiệu mạch đập cổ tay …………….……… ……….…………………………………………………………… …………….……… ……….…………………………………………………………… Ngày giao đồ án: 15/03/2023 Ngày nộp đồ án: 15/07/2023 Ngơn ngữ trình bày: Bản báo cáo: Tiếng Anh Tiếng Việt Trình bày bảo vệ: Tiếng Anh Tiếng Việt TRƯỞNG KHOA TRƯỞNG BỘ MÔN GIẢNG VIÊN HƯỚNG DẪN (Ký, ghi rõ họ tên) (Ký, ghi rõ họ tên) (Ký, ghi rõ họ tên) i COMMITMENT • Project: Development of a Wearable AI-based Device for Wrist Pulse Diagnosis • Lecturer: Bui Ha Duc, Ph.D • Student: Dao Thanh Quan • Student ID: 19134081 - Class: 19134 • Adress: Phuoc Long B Ward, Thu Duc City, Ho Chi Minh City • Phone number: 0962600379 • Email: quandaoforwork@gmail.com • Student: Le Quoc Tuan • Student ID: 19134091 - Class: 19134 • Adress: Linh Chieu Ward, Thu Duc City, Ho Chi Minh City • Phone number: 0935920345 • Email: lequoctuan.fwork@gmail.com • Graduation thesis submission date: n • Commitment: “I affirm that the graduation thesis presented here is the result of my research and efforts I have not replicated any content from published articles without appropriate citations Should any breach be identified, I acknowledge full accountability for the consequences.” Ho Chi Minh City, … July 2023 ii ACKNOWLEDGMENT First and foremost, on behalf of my team, I would like to extend my heartfelt thanks to our supervisor, Bui Ha Duc Ph.D Your unwavering commitment, expertise, and guidance have been instrumental in shaping my research and steering me toward success Your mentorship, patience, and constant encouragement have played a significant role in shaping my academic journey Your profound knowledge and invaluable insights have helped me overcome challenges and broaden my intellectual horizons I am truly grateful for your dedication and support I would also like to extend my sincere appreciation to Ho Chi Minh City University of Technology and Education for providing me with an exceptional learning environment and resources The university has been a constant source of inspiration, fostering an atmosphere of growth, innovation, and academic excellence The diverse faculty and staff have been instrumental in shaping my academic and personal development, and I am indebted to them for their commitment to nurturing future scholars and leaders My deepest gratitude goes to my family for their unwavering love, encouragement, and n understanding Your support has been the foundation of my journey, and I am forever grateful for your sacrifices, belief in me, and the countless ways you have cheered me on Your unwavering support has provided me with the strength and motivation to overcome challenges and strive for excellence Lastly, I would like to express my gratitude to all those who have contributed to my growth and development, directly or indirectly Your encouragement, advice, and belief in my abilities have been invaluable throughout this challenging yet rewarding journey Sincerely, Dao Thanh Quan iii ABSTRACT Pulse diagnosis, an integral part of traditional medicine (TM), is a crucial diagnostic method alongside looking, listening, and asking It involves the practitioner placing their three fingers on the patient's radial artery at the wrist to analyze their health condition This method has been used for thousands of years in TM and continues to be highly regarded for its convenience, affordability, and non-invasive nature Pulse diagnosis remains a strong contender for disease diagnosis even in modern times Recent studies have highlighted the significance of the wrist pulse signal as a bloodstream signal that can provide valuable insights for disease analysis However, traditional pulse diagnosis (TPD) heavily relies on the expertise of practitioners The measurement and interpretation involved in TPD typically demand years of training for practitioners Additionally, there is a lack of standardized communication and sharing of pulse signal experiences among different practitioners These factors present challenges in the further development and widespread application of TPD in modern clinical practice n Advancements in sensor technologies, signal processing, and pattern recognition have paved the way for significant advancements in the computational analysis of pulse signals These developments have led to the creation of three types of sensors for pulse signal acquisition: pressure sensors, photoelectric sensors, and ultrasonic sensors These sensors enable the simulation of pulse signal analysis, resembling the expertise of practitioners Signal processing and pattern recognition methods have been devised to interpret and analyze pulse signals As a result, pulse signals have been extensively investigated for various applications, including pulse waveform classification, prediction, and the diagnosis of numerous diseases such as cholecystitis, nephrotic syndrome, diabetes, and more The objectives of this study are to research, design, and fabricate a measuring device equipped with a piezo sensor to capture wrist pulse signals, and apply artificial intelligence to analyze the recorded signal In addition, we have gathered data on wrist pulse (using our developed device) and blood glucose levels (from a commercially available device) from a group of individuals over several days, with measurements taken at various times throughout the day By applying digital signal processing techniques, we have effectively eliminated noise sources such as high-frequency noise, electrostatic noise, and baseline wander from the collected data Our ultimate goal is to leverage this processed data to create an AI model capable of predicting pulse signals and blood glucose levels Our research findings have been accepted and approved at the 2023 International Conference on System Science and Engineering (ICSSE) iv TABLE OF CONTENTS NHIỆM VỤ ĐỒ ÁN TỐT NGHIỆP i COMMITMENT ii ACKNOWLEDGMENT iii ABSTRACT iv TABLE OF CONTENTS v LIST OF TABLES viii LIST OF FIGURES ix LIST OF ACRONYMS xii CHAPTER 1.INTRODUCTION 1.1 Motivation 1.2 Scientific and practical significances 1.3 Objectives 1.4 Research methods 1.5 Structure of the report CHAPTER 2.LITERATURE REVIEW Introduction to Traditional Medicine 2.2 Introduction to the use of sensors in collecting pulse wave signals n 2.1 2.2.1 Photoplethysmography (PPG) sensors 2.2.2 Piezoresistive sensors 2.2.3 Capacitive pressure sensors 11 2.2.4 Piezoelectricity sensors 13 2.3 Research overview 15 2.4 Introduction to noise-removing technique used for pulse signal 17 2.4.1 Finite Impulse Response (FIR) filter 18 2.4.2 Infinite Impulse Response (IIR) filter 19 2.4.3 Wavelet-based filter 20 2.5 Introduction to metrics used in time series 21 2.5.1 Introduction to signal-to-noise ratio 21 2.5.2 Introduction to mean squared error 22 2.6 Introduction to Machine Learning in Traditional Medicine 22 2.6.1 Application of machine learning 22 2.6.2 Introduction to statistical model used in time series forecasting 23 v 2.6.3 Introduction to deep learning model used in time series forecasting 25 CHAPTER 3.DESIGN MECHANICAL SYSTEM 29 3.1 Hardware objectives 29 3.2 Technical requirements 29 3.3 Design options 30 3.3.1 Design the bracelet 30 3.3.2 Design the electrical box 38 CHAPTER 4.DESIGN ELECTRONICS – CONTROL SYSTEM 40 4.1 Electronics – control system’s objectives 40 4.2 Technical requirements 40 4.3 Design options 41 4.3.1 Cetral control block 41 4.3.2 Power supply block 41 4.3.3 Sensor block 42 4.3.4 Signal processing block 42 4.4 Hardware experiment 48 4.4.1 Experiment set up 48 n 4.4.2 Experimental result 50 CHAPTER 5.DESIGN PULSE WAVE FORECASTING ALGORITHM 53 5.1 Objectives 53 5.2 Data Preparation 53 5.2.1 Equipment 53 5.2.2 Patient recruitment and protocol 53 5.3 Preprocessing algorithm 56 5.3.1 Applying wavelet transform in filtering signal 56 5.3.2 Baseline wander removal 59 5.4 Time series forecasting model 62 5.4.1 Time2Vec 63 5.4.2 Gating mechanism 63 5.4.3 Feature selection network 66 5.4.4 Interpretable multi-head attention 67 5.4.5 Locality enhancement 68 5.5 Experiment 68 5.5.1 Training procedure 68 vi 5.5.1 Experimental result 69 CHAPTER 6.BLOOD GLUCOSE MEASUREMENT 72 6.1 Objectives 72 6.2 Blood glucose measuring model 73 6.3 Experiment 75 6.3.1 Dataset preparation 75 6.2.5 Training procedure: Blood glucose measuring model 76 6.2.6 Experimental result: Blood glucose measuring model 77 CONCLUSION AND DISCUSSION 78 REFERENCE 79 APPENDIX I n vii We define the input time series as 𝑋𝑡𝑖 where 𝑖 denotes the variable in the dataset and t is number of time steps Given a sequence with length 𝑇: (𝑥1𝑖 , 𝑥𝑇𝑖 ) , we would like to measure the value of blood glucose level 𝑦̂ n Figure 6.3: Block diagram of blood glucose measuring model First of all, the input sequence is process by a temporal convolutional network It is a fundamental operation used in time series forecasting By applying a convolution kernel to a time series sequence, it enables the model to capture local patterns and dependencies within the data In other words, temporal convolutional network acts as a sliding filter across the input sequence This filter can capture local patterns by performing element-wise multiplication and summing 74 the results Hence, the model can extract relevant features from the sequence, recognize patterns at different time scales Secondly, the process data is fed to a LSTM-based encoder-decoder While the encoder extracts and summarizes the information of the data, the decoder generates the output based on the summarized information Besides, LSTM has proven its ability to capture long-term dependencies and detect temporal pattern, so using it help the model deal with long sequence of data Additionally, LSTM can alleviate the effect of exploding and vanish gradient After that, a skip connection to add the original input into the output of the decoder Skip connection help the model have a direct access to earlier time steps, allowing it to capture long-term dependencies more effectively Moreover, it can improve the gradient flow of the model by allowing the gradient to propagate directly to earlier layers, facilitating better information flow and faster convergence A dropout layer is used after the skip connection Finally, we apply an additional non-linear process layer to the output of dropout layer and then another dense layer for the output 6.3 Experiment 6.3.1 Dataset preparation n In this application, we also use the dataset that we mentioned in chapter 5.2 In particular, the piezo signal is the inputs of that model, the equivalent blood glucose level is the label However, we aware that our dataset is not large enough, so we use an additional dataset called D1NAMO [2] to pre-train our model Overall, D1NAMO dataset is an open multi-modal dataset consisting of ECG, breathing, accelerometer signals, blood glucose level measurement and annotated food in real-life condition It is acquired on 20 healthy subjects and patients with type-1 diabetes The acquisitions are conducted by subjects themselves by starting the sensor after waking up and shutting it down before going to bed The process lasts days per subject For blood glucose measurement, healthy subjects were asked to measure their blood glucose level times per day using the Bayer Contour XT glucose meter with Bayer Next strips, while diabetes patients measure theirs using iPro Processional CGM sensor For the sake of pre-training, we only use ECG, and blood glucose data in this dataset Because ECG records the activity of the heart so it relates to our wrist pulse wave signal to some extent The ECG data is taken with a 1-lead sensor, including two silver-coated nylon electrodes at the sample rate of 250 Hz However, the acquisition process of this data is conducted by each subject so there are many problems with the result such as poor electrode contact or the electrode 75 is not connected, etc Hence, the data is extremely noisy and we have to manually check and discard the failure data The remaining data is then applied wavelet filter, baseline wander removal, scaling as our data Additionally, it is resampled to the same sample rate with our data n Figure 6.4: Zephyr Bioharness device used in D1NAMO project (Source: [2]) Additionally, the blood glucose level in this dataset is measured in term of a molar concentration or mmol/L, so we convert it into the same unit with our device as the formular below: 18 𝑚𝑚𝑜𝑙 𝑚𝑔 =1 𝐿 𝑑𝐿 (6.1) 6.2.5 Training procedure: Blood glucose measuring model First of all, for two prepared datasets, we also divide into training, validation, and test datasets according to the ratio of 9-1-1 as we in the forecasting task Besides, we also conduct random searching to find the optimal hyperparameters for the model for 60 iterations We then divide each dataset into windows to be prepared for training Similar to the training process for forecasting model, we also pick 300000 windows for the training dataset and 30000 for the validation dataset randomly for each training iterations Below is the list of optimal hyperparameters: 76 • Hidden state size: 64 • Dropout rate: 0.1 • Learning rate: 0.004 • Max gradient norm: 0.01 The fixed parameters for this model are listed below: • Input length: 128 (2 seconds) • Output length: • Minibatch size: 64 • Early stopping patience: 10 • Epochs: 100 • Iterations: 50 • Loss function: Mean Square Error The training and inference procedure for this model is operated on NVIDIA T4 GPU on Google Colab The time it takes for training process is a whole day 6.2.6 Experimental result: Blood glucose measuring model n According to the objectives, the performance of the blood glucose measuring model is evaluated by two different metrics, which are mean absolute error and mean absolute percentage error In particular, we divided the test dataset into two subsets with including one with blood glucose level within under 100 mg/dL and the other with blood glucose level within 100 mg/dL or more For the lower subset, we use MAE to evaluate, and mean absolute percentage error (MAPE) is used for the higher one The formula of MAPE is defined as follow: 𝑁 |𝑦̂𝑖 − 𝑦𝑖 | 𝑀𝐴𝑃𝐸 = ∑ ∗ 100% 𝑁 𝑦𝑖 (6.2) 𝑖=1 where N is the number of blood glucose values, 𝑦̂𝑖 and 𝑦𝑖 are predicted and actual blood glucose level, respectively The results of the evaluation process are as follow: • 𝐺𝑙𝑢𝑐𝑜𝑠𝑒 𝐶𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 < 100 • 𝐺𝑙𝑢𝑐𝑜𝑠𝑒 𝐶𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 ≥ 100 𝑚𝑔 𝑑𝐿 𝑚𝑔 𝑑𝐿 : 𝑀𝐴𝐸 = 2.48 : 𝑀𝐴𝑃𝐸 = 3% Although our model currently doesn’t achieve a state-of-the-art performance, it still satisfies the objectives that we propose, so the current performance is acceptable However, it can be improved in the future by adding more data as well as optimizing the architecture and hyperparameters 77 CONCLUSION AND DISCUSSION Overall, we can finish all objectives that we proposed We built a wearable device that can precisely record the pulse signal from the wrist of a subject The precision of it has already been evaluated by a trustworthy ECG sensor Moreover, we can build a deep learning model that can forecast the pulse wave signal seconds into the future and another one that can use the forecasted data to measure the blood glucose level However, our project cannot be without shortage One of which is our dataset We aware that our data is not large as well diverse enough It also doesn’t contain data from diabetes patients This is because we cannot recruit patient for the data acquisition purpose Besides, we currently don’t have an architecture that is large enough to build a model that can forecast longer Hence, our future works are: • Acquire a larger and more diverse dataset including data from normal subjects, diabetes patients of different ages as well as different compilations of diabetes • Increase the scale of model to make it able to forecast longer • Make use of the interpretability properties of the attention mechanism of the forecasting n model, and apply interpretability to the blood glucose measuring model to provide insights about the decisions of the models 78 REFERENCE [1] Chuanglu Chen et al, A 3D Wrist Pulse Signal Acquisition System for Width Information of Pulse Wave, Sensors, pp 1-16, vol 20, MDPI, 2019 [2] Fabien Dubosson et al, The open D1NAMO dataset: A multi-modal dataset for research on non-invasive type diabetes management, Informatics in Medicine Unlocked, pp 1-9, vol 13, Elsevier, 2018 [3] Cuiwei Li et al, Detection of ECG characteristic points using wavelet transforms, IEEE Transactions on Biomedical Engineering, pp 1-8, vol 42, IEEE, 1995 [4] Timibloudi Enamamu at el, Continuous m-Health Data Authentication Using Wavelet Decomposition for Feature Extraction, Sensors, pp 1-22, vol 20, MDPI, 2020 [5] Arthur de Sá Ferreira, Resonance phenomenon during wrist pulse-taking: A stochastic simulation, model-based study of the ‘pressing with one finger’ technique, Biomedical Signal Processing and Control, pp 1-2, vol 8, Elsevier, 2013 [6] Tiantian Guo et al, A Review of Wavelet Analysis and Its Applications: Challenges and n Opportunities, IEEE Access, pp 1-35, vol 10, IEEE, 2022 [7] International Diabetes Federation, IDF Diabetes Atlas, Belgium: International Diabetes Federation, Brussels, 2021 [8] Ming-Yie Jan et al, The Physical Conditions of Different Organs Are Reflected Specifically in the Pressure Pulse Spectrum of the Peripheral Artery, Cardiovascular Engineering: An International Journal, pp 1-2, vol 3, Springer, 2023 [9] Nan Li et al, The Correlation Study of Cun, Guan and Chi Position Based on Wrist Pulse Characteristics, IEEE Access, pp 1-13, vol 9, IEEE, 2021 [10] Ganesh R Naik, Computational Intelligence in Electromyography Analysis – A Perspective on Current Applications and Future Challenges, InTech, 2012 [11] Shi Zhen Li, Shih-chen Li, Pulse Diagnosis, Paradigm Press, 1993 [12] Jiapu Pan, Willis J Tompkins, A Real-Time QRS Detection Algorithm, IEEE Transactions on Biomedical Engineering, pp 2-3, vol BME-32, IEEE, 1985 [13] Salvador Quiroz-González et al, Acupuncture Points and Their Relationship with Multireceptive Fields of Neurons, Journal of Accupunture and Meridian Studies, pp 29, vol 10, Elsevier, 2017 [14] Gunther Schemelzeisen-Redeker et al, Time Delay of CGM Sensors: Relevance, Causes, and Countermeasures, Journal of Diabetes Science and Technology, pp 1-2, vol 9, Sage, 2015 79 [15] Vijayakumari et al, Analysis of noise removal in ECG signal using symlet wavelet, International Conference on Computing Technologies and Intelligent Data Engineering, pp 1-6, IEEE, 2016 [16] Tanuj Yadav, Rajesh Mehra, Denoising ECG Signal Using Daubechies and Symlet Wavelet Transform Techniques, International Journal of Advanced Research in Computer and Communication Engineering, pp 1-6, vol 5, 2016 [17] Yu-Fend Chung, New vision of the pulse conditions using Bi-Sensing Pulse Diagnosis Instrument, International Conference on Orange Technologies, pp 1-4, IEEE, 2013 [18] Changbo Zhao et al, Advances in Patient Classification for Traditional Chinese Medicine: A Machine Learning Perspective, Evidence-Based Complementary and Alternative Medicine, vol 2015, pp 1-19, vol 2015, Hindawi, 2015 [19] Djork-Arné Clevert, Thomas Unterthiner, and Sepp Hochreiter, Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs), link https://arxiv.org/abs/1511.07289v5, 6/2023 n [20] Seyed Mehran Kazemi et al, Time2Vec: Learning a Vector Representation of Time, link https://arxiv.org/abs/1907.05321, 6/2023 [21] Bryan Lim at el, Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting, link https://arxiv.org/abs/1912.09363, 4/2023 [22] Ashish Vaswani et al, Attention Is All You Need, link https://arxiv.org/abs/1706.03762, 4/2023 80 APPENDIX n I Proceedings of 2023 International Conference on System Science and Engineering (ICSSE) Development of a Wearable Human Pulse Measuring Device using Piezoelectric Sensor Quan Dao Thanh Department of Mechanical Engineering HCMC University of Technology and Education Ho Chi Minh City quandaoforwork@gmail.com Tuan Le Quoc Department of Mechanical Engineering HCMC University of Technology and Education Ho Chi Minh City lequoctuan.fwork@gmail.com n Abstract— Pulse diagnosis, an ancient technique in traditional Chinese medicine (TCM), is an effective non-invasive method for diagnosing human diseases Traditionally, the practitioners use their fingertips to sense the pulse of the blood vessels on the patient's wrist However, this method heavily and subjectively depends on the doctor's experience Additionally, it’s impossible to store the patient's pulse records for posterior review or analysis To solve such restrictions, a wearable device that could precisely measure the wrist-based pulse signal and save it as digital data was proposed In the proposed device, a piezoelectric sensor, which is a highly sensitive, lightweight, stable sensor, was used to record variations of the pulse Since wrist pulse signals collected by the device contain various types of noises which make it difficult to assess the shape of the pulse wave, a slight low-pass filter was applied to remove highfrequency noises The processed data was then compared with simultaneous electrocardiogram (ECG) data for the sake of evaluation The preliminary results showed that recorded data from the proposed device is consistent with the equivalent ECG data Keywords—piezoelectric sensor, traditional Chinese medicine, wearable device, pulse wave analysis, Arterial pulse I INTRODUCTION For thousands of years, pulse diagnosis has played an important role in traditional Chinese medicine (TCM) for disease analysis This is a well-known noninvasive assessment approach that is successfully used not only in China but also in many Eastern countries Compared with ECG signals which can be used to monitor the health status of the human heart, wrist pulse waves can reflect the resonance oscillation between the heart and other internal organs through the arterial system Hence, it can be used to get even more information about the human body’s health status besides the heart In practice, doctors would place the matching fingers on the patient's wrist at three places which are cun, guan, and chi They then feel the characteristics of the pulse, such as the pulse rate, trend, strength, length, width, etc to make diagnoses for the patient However, Chinese medicine practitioners may provide varied answers because the accuracy of the diagnosis depends subjectively on their intuition and experience Recently, by taking advantage of the developed combination of digital signal processing technology and biomedical knowledge, TCM pulse measuring devices can now visualize and analyze the measured pulse profile instantaneously [1]–[3] Duc Bui Ha Ph.D Department of Mechanical Engineering HCMC University of Technology and Education Ho Chi Minh City ducbh@hcmute.edu.vn There have been many different approaches to describe the pulse capture technique using different kinds of sensors For example, John N Lygouras et al [4], Kun-Chan Lan et al [5], and Leonid S Lovinsky [6] use a photoplethysmography sensor to measure pulse signals at the fingertips and wrist Although they achieved effective outcomes with their device, their processes didn’t follow the nature of wrist pulse diagnosis in traditional Chinese medicine While a photoplethysmography sensor uses infrared light to measure the volumetric changes in blood, it cannot obtain the pulse strength as well as the resonance oscillation between the heart and other organs Lisheng Xu et al [7] implemented a wavelet-based cascaded adaptive filter (CAF) to remove the baseline wander of the measured signal from a single piezoelectric sensor, this study gives many positive results about piezoelectric data acquisition as well as noise filtering from the signal, but using a single sensor leads to the loss of important data characteristics and misrepresents the genuine function of pulse capture in TCM In some other studies, Chung et al [8] used three sensors to get the wrist’s pulse information simultaneously and used Three-Dimension Pulse Mapping (3DPM) to simulate the actual touch sensation, or Chuangly Chen et al [9] used a sensor array composed of rows x columns MEMS sensors to record wrist pulse pressure and mapped it to 3D data By obtaining 3D pulse wave data, it is certain that they can extract much more important features from the pulse such as depth, and length However, it is also a huge disadvantage which is the significantly increasing computational weight for the posterior analysis Additionally, their signal acquisition systems are clunky and expensive so it is difficult to develop them into a wearable device To address the limitations of these prior studies [4]–[9], we propose a wearable sensor system including a 3x1 piezoelectric sensor array to record wrist pulse waves at cun, guan, and chi positions Our contribution to this work includes designing and implementing a wearable piezoelectric-based device Furthermore, the recorded data is saved in comma-separated values format for the latter processing and analyzing format The remainder of this paper is organized as follows: Section presents our proposed methodology including data acquisition, and data analysis The recorded results are presented in Section and the paper is finally concluded in Section Proceedings of 2023 International Conference on System Science and Engineering (ICSSE) II METHODOLOGY A Wrist pulse wave In Traditional Chinese Medicine, there are four diagnostic methods used to examine human health conditions namely inspection, auscultation, olfaction and interrogation, and palpation Each method is based on different information such as external conditions, family history, diets, living states, etc of the patients to examine their wellness [10] In this study, we mainly focus on wrist pulse wave diagnosis which belongs to palpation methods The wrist pulse diagnosis is an important method to assess a person’s overall health conditions and identify imbalances within the human body It is commonly believed by TCM practitioners that wrist pulses can reflect the flow of qi, which is a vital energy or life force, and everything in the world is made up of qi Qi circulates throughout the human body as well as every organ inside it through meridian channels This flow is then reflected in many acupuncture points in the body including cun, guan, chi positions Hence, wrist pulse diagnosis is a way to capture the information of qi through acupuncture points According to TCM theory, the stability and balance of the flow of qi are the most important concepts used to judge the wellness of the human body By assessing the flow of qi through the wrist pulse, TCM practitioners can grasp the whole information about the internal organs of a person [11], [12] Moreover, many recent researches have given scientific proof of this kind of diagnosis In detail, during the respiratory process, the heart creates an oscillation that is distributed throughout the arterial system of the human body Side-branch organs then react to the oscillation, generate harmonic forces with maximum amplitudes near their own natural frequencies, and then distribute back to the arterial system Based on this “frequency match” theory, it is recorded that the frequencies at several acupuncture points match the natural frequencies of several organs [13], [14] To assess the wrist pulse, TCM practitioners simultaneously place their index, middle and ring fingers on the patient’s radial artery, which correspond to cun, guan, and chi, respectively These positions are commonly believed to represent the states of internal organs of the human body For example, cun, guan, and chi positions on the left hand correspond to heart, liver, kidney while those on the right hand correspond to lung, stomach and kidney, respectively Because the pulse varies according to applied static forces, practitioners then respectively apply light, medium, heavy forces to feel for various qualities, including pulse rate, trend, strength, length, width, etc Based on these qualities, practitioners with experiences can tell the imbalance and overall health state of a person [9], [15], [16] a heartbeat Piezoelectric sensors also have a fast response time, which is important for accurately capturing the pulse wave The pulse wave is a dynamic signal that changes rapidly, and a fast response time is needed to capture these changes Photoplethysmography sensor also has sensitive characteristic and fast response time, but this type of sensor cannot precisely reflect the nature of TCM techniques which depend on the forces applied to the vessels at the wrist Because of the adjacent of cun - guan - chi position, the sensor's size must be comparable to the size of a typical person's finger In this study, we use piezoelectric sensors (ID: DT1-052K) developed by TE Connectivity Inc with size 16 mm x 41 mm (width x height) and the total thickness is 40 μm Fig Piezoelectric sensor n C Hardware design The hardware architecture of the proposed system was shown in Fig It was divided into three parts: a device including three Piezoelectric sensors, ADC and Raspberry Pi for data acquisition and analysis, computer for data visualization and storage When utilizing a piezoelectric sensor to measure the pulse signal of the wrist, the primary challenge is the presence of electrostatics on the body The direct contact of the sensor with the skin causes the electrostatic to interfere with the signal, resulting in signal inaccuracy and significant noise As a result, the received signal is not reliable The final design of the system is shown in Fig In order to avoid electrostatic interference and false signals when the sensor is in direct contact with the skin, an isolator element is incorporated into the sensor design, enabling indirect measurement without direct skin contact The isolator (Fig 3b) elements can rotate along the axis, allowing us to apply different forces on them The sensor tips represent the fingers that make contact with the wrist and transmit the vibrations to the piezo sensor It is free to vibrate thanks to the hollow space in the isolator B Piezoelectric sensor Piezoelectric sensors are electronic devices that can transform mechanical or thermal energy into electrical signals, utilizing the principle of electromechanical coupling This is based on the phenomenon of piezoelectricity, whereby certain materials generate an electrical voltage when exposed to mechanical stress, and conversely, produce mechanical stress when subjected to an electrical voltage This type of sensor has several specific advantages over other types of pressure sensors that make it well-suited for TCM Firstly, piezoelectric sensors are highly sensitive and can detect the pulse wave which is a very small pressure wave generated by Fig The hardware architecture of the proposed system Proceedings of 2023 International Conference on System Science and Engineering (ICSSE) Fig (Left) Proposed design of the system (Right) Vertical half-view of the isolator According to [15], the distance between Guan and Cun is about 10 mm, and Chi is about 10-15mm proximal from Cun Hence, the distance between each sensor tip is 10 mm A bracelet was designed to fit the wrist size of an adult Three M6 bolts were placed on the upper part of the bracelet to apply force to the isolators E Data preprocessing D Data acquisition n Based on the nature of piezoelectric sensors which generate voltage with respect to the applied pressure, an additional analog-to-digital converter (ADC) is needed to digitize the generated voltage signal Additionally, the human wrist can only generate a very small pressure, and it was tested that the voltage generated by piezoelectric sensors is quite small Hence, a high-resolution ADC must be used to record even the slightest changes in piezoelectric signal In this study, a 32-bit ADC named ADS1263 was chosen to satisfy the aforementioned requirements To read data from this ADC, an additional Raspberry Pi Zero which was directly connected to the ADS1263 was used Furthermore, for the sake of evaluation, an ECG sensor, which is a 3-wire ECG named SEN0213 (DFRobot), was recorded simultaneously However, the ECG sensor was recorded separately by an Arduino Uno microcontroller board instead of a Raspberry Pi Zero An additional circuit was built to make sure that two types of sensors are recorded simultaneously In this circuit, a button was used to decide when to start recording After the button is clicked, Raspberry Pi Zero sends a signal to Arduino Uno through an optocoupler and then two boards start to record the signal simultaneously As mentioned before, the recorded data is contaminated by several types of noises such as 50 Hz frequency noise created by the power supply, or baseline wander Besides, because of the nature of piezoelectric sensors, it must be exposed to the human wrist to record the pressure signal Hence, the recorded signal is directly affected by electrostatic noise as well as muscular noise happening on human skin [17] As a result, the validation process cannot be done without filtering these noises However, because of the objective of this study, only high-frequency noises are focused Specifically, the recorded data is put through a lowpass filter with a cutoff frequency of 20 Hz to partially remove high-frequency data Furthermore, in order to decrease the delay offered by the filter, a Butterworth filter is proposed in this study III RESULT A Hardware The system comprises a bracelet and a control box, as shown in Fig Both of which are fabricated using 3D printing technology and PLA material with a LayerHeight of 0.2 mm and Infill of 100% This process facilitates precision machining at a low cost The control box, with dimensions of 150 mm x 100 mm x 50 mm, is compact and can be wired electrically It also features a cooling fan that aids in dissipating heat generated Fig The electrical circuit used for validation process Fig Actual image of the system during operation Proceedings of 2023 International Conference on System Science and Engineering (ICSSE) during operation The entire system weighs less than 500 g, making it easily portable B Data acquisition and preprocessing The experimental process was conducted by recording ECG signals and piezoelectric sensors simultaneously The raw data of this process was shown in Fig Specifically, the recorded ECG data was shown in the first chart, while the other three charts are piezoelectric signals measured at cun, guan, chi positions, respectively As mentioned before, due to the small pressure generated by blood vessels in the human wrist, the obtained data from piezo sensor is extremely small which is in the range from 0.0175 to 0.0175 (V) At the first glance, both types of signal are highly affected by noises that make it hard it assess the recorded data However, by taking a closer look at the red circles, it is noticeable that the signal generated by piezoelectric sensors has a consistent beat over time which aligns with the ECG data Moreover, after applying the butterworth low-pass filter, noisy data is significantly removed in all four recorded data Although the filtered data still contains noises, it is much easier for the validation process Fig contains 5-second recorded signal on different subjects in which the order of the chart is similar to one mentioned in the raw data figure Similar to what is shown in the raw data figure, red circles are draw on the heart beat position It can be seen the heart beats recorded by our device are consistent with those by ECG signal In conclusion, it is shown by the preliminary result that the data gained from the proposed device is consistent with the equivalent ECG data n Fig The 4-second recorded data of ECG sensor and the proposed device Proceedings of 2023 International Conference on System Science and Engineering (ICSSE) n Fig The 5-second filtered data of ECG sensor and the proposed device on different objects IV ACKNOWLEDGMENT DISCUSSION First of all, we would like to give a special thanks to Ho Chi Minh City University of Technology and Education for giving us an opportunity to this study Besides, this study cannot be completed without a devoted support from our supervisor, and we are very grateful to him In this study, we proposed a wearable device that utilizes a piezoelectric sensor to precisely measure and record wristbased pulse signals in a non-invasive manner The device is designed to overcome the limitations of traditional pulse diagnosis, which heavily relies on the practitioner's experience and cannot store patient records for later analysis To validate the effectiveness of the proposed device, the recorded data is compared with simultaneous electrocardiogram (ECG) data, and the preliminary results demonstrate consistency between the two data sets We also highlight the advantages of their approach, which provides more information about the resonance oscillation between the heart and other internal organs through the arterial system than ECG signals alone Overall, this study provides a promising solution to improve the accuracy and objectivity of traditional pulse diagnosis in TCM and could potentially lead to more personalized and effective treatments REFERENCE [1] [2] [3] D Jia et al., “A Fiber Bragg Grating Sensor for Radial Artery Pulse Waveform Measurement,” IEEE Trans Biomed Eng., vol 65, no 4, pp 839–846, Apr 2018, doi: 10.1109/TBME.2017.2722008 J Gong, S Lu, R Wang, and L Cui, “PDhms: Pulse Diagnosis via Wearable Healthcare Sensor Network,” in 2011 IEEE International Conference on Communications (ICC), Kyoto, Japan: IEEE, Jun 2011, pp 1–5 doi: 10.1109/icc.2011.5963341 C Jin, C Xia, S Zhang, L Wang, Y Wang, and H Yan, “A Wearable Combined Wrist Pulse Measurement Proceedings of 2023 International Conference on System Science and Engineering (ICSSE) [4] [5] [6] [7] [8] [9] [11] [12] [13] [14] [15] [16] pp 81–89, Apr 2017, doi: 10.1016/j.jams.2017.01.006 [17] A C Marshall, “Traditional Chinese Medicine and Clinical Pharmacology,” in Drug Discovery and Evaluation: Methods in Clinical Pharmacology, F J Hock and M R Gralinski, Eds., Cham: Springer International Publishing, 2020, pp 455–482 doi: 10.1007/978-3-319-68864-0_60 n [10] System Using Airbags for Pressurization,” Sensors, vol 19, no 2, p 386, Jan 2019, doi: 10.3390/s19020386 J N Lygouras and P G Tsalides, “Optical-fiber finger photo-plethysmograph using digital techniques,” IEEE Sens J., vol 2, no 1, pp 20–25, Feb 2002, doi: 10.1109/7361.987057 K.-C Lan, G Litscher, and T.-H Hung, “Traditional Chinese Medicine Pulse Diagnosis on a Smartphone Using Skin Impedance at Acupoints: A Feasibility Study,” Sensors, vol 20, no 16, p 4618, Aug 2020, doi: 10.3390/s20164618 L S Lovinsky, “Urgent Problems of Metrological Assurance of Optical Pulse Oximetry,” IEEE Trans Instrum Meas., vol 55, no 3, pp 869–875, Jun 2006, doi: 10.1109/TIM.2006.873784 L Xu, D Zhang, K Wang, N Li, and X Wang, “Baseline wander correction in pulse waveforms using wavelet-based cascaded adaptive filter,” Comput Biol Med., vol 37, no 5, pp 716–731, May 2007, doi: 10.1016/j.compbiomed.2006.06.014 Yu-Feng Chung, Yu-Wen Chu, Cheng-Ying Chung, Chung-Shing Hu, Ching-Hsing Luo, and Cheng-Chang Yeh, “New vision of the pulse conditions using BiSensing Pulse Diagnosis Instrument,” in 2013 1st International Conference on Orange Technologies (ICOT), Tainan: IEEE, Mar 2013, pp 5–8 doi: 10.1109/ICOT.2013.6521143 C Chen, Z Li, Y Zhang, S Zhang, J Hou, and H Zhang, “A 3D Wrist Pulse Signal Acquisition System for Width Information of Pulse Wave,” Sensors, vol 20, no 1, p 11, Dec 2019, doi: 10.3390/s20010011 N Li, J Yu, H Hu, X Mao, Y Zhao, and L Huang, “The Correlation Study of Cun, Guan and Chi Position Based on Wrist Pulse Characteristics,” IEEE Access, vol 9, pp 28917–28929, 2021, doi: 10.1109/ACCESS.2021.3058791 A O., A B., S J., and P J., “EMG Decomposition and Artefact Removal,” in Computational Intelligence in Electromyography Analysis - A Perspective on Current Applications and Future Challenges, G R Naik, Ed., InTech, 2012 doi: 10.5772/50819 Shi Zhen Li and Shih-chen Li, Pulse Diagnosis A D S Ferreira, “Resonance phenomenon during wrist pulse-taking: A stochastic simulation, modelbased study of the ‘pressing with one finger’ technique,” Biomed Signal Process Control, vol 8, no 3, pp 229–236, May 2013, doi: 10.1016/j.bspc.2012.10.004 M.-Y Jan, H Hsiu, T.-L Hsu, W.-K Wang, and Y.-Y L Wang, “The Physical Conditions of Different Organs Are Reflected Specifically in the Pressure Pulse Spectrum of the Peripheral Artery,” 2003 C Zhao, G.-Z Li, C Wang, and J Niu, “Advances in Patient Classification for Traditional Chinese Medicine: A Machine Learning Perspective,” Evid Based Complement Alternat Med., vol 2015, pp 1– 18, 2015, doi: 10.1155/2015/376716 S Quiroz-González, S Torres-Castillo, R E LópezGómez, and I Jiménez Estrada, “Acupuncture Points and Their Relationship with Multireceptive Fields of Neurons,” J Acupunct Meridian Stud., vol 10, no 2, S K L 0 n