INTRODUCTION
Motivation
Technological advancements are transforming the sports industry globally, with rapid growth observed in many countries Innovations in equipment and devices are providing significant advantages across various sports, including football.
In October 2015, FIFA announced the approval of Electronic Performance and Tracking System (EPTS) devices during its 129th Annual General Meeting, enabling football players and coaches to utilize integrated technology in official competitions This decision sparked a significant rise in the adoption of smart devices in football globally, particularly in the UK, where the Premier League, the world's most renowned football tournament, has seen various clubs like Manchester City, Newcastle United, and Fulham incorporate wearables into their training and performance analysis.
Figure 1.1.a The Cityplay tracker sensors and case [7]
Since 2019, the Vietnam National Football Team has utilized wearable devices, which played a crucial role in their victory over Indonesia during the King's Cup 2019 tournament These innovative devices have enhanced the team's performance and strategy on the field.
Vietnamese athletes wore sports bras integrated with a GPS module that enabled real-time location tracking, endurance monitoring, and heart rate and speed calculations This innovative technology assisted coaches in tailoring training programs to meet the specific needs of each player, enhancing the effectiveness of training sessions for the Vietnamese team.
Figure 1.1.b The Vietnamese team uses the GPS shirt of Catapult [9]
The Vietnamese National Football team has made significant progress with the integration of technological equipment; however, a major limitation hindering further improvement is the high cost of these devices The Catapult sport bra, primarily used by the national team, requires a membership subscription, costing $179 per device kit annually, totaling over $4,000 for the team.
To significantly advance the country's football, it is essential to develop a device system akin to the Catapult Bra, but at a more affordable price Crucially, this system must be mastered with local technology, free from foreign interference.
Objectives
Based on above-mentioned issues, our group has implemented a project named Sport Performance Tracking System to help solving these problems in the most effective way, including:
- Selecting the most appropriate sensors due to their specifications
- Processing input signal including physiological and technological data
- Hardware construction due to problem requirements
- Multi-device connection using Bluetooth Low Energy mesh
- Results and evaluation demonstration for analysis
- Server and database usage for data storing.
Scope
Given our technical, financial, and knowledge constraints, along with time limitations, we opted to create a prototype for this thesis This approach enables others to explore and develop future iterations of the project.
Thesis organization
This thesis comprises of 7 chapters, which are fluently and logically organized:
- Chapter 1: The introduction of project and approaching methods
- Chapter 2: Literature review of the related fundamental principles
- Chapter 3: Hardware design and implementation of device system
- Chapter 4: Controlling software and mathematic calculating algorithms installed in the system
- Chapter 5: Building the wireless connection among several devices using Bluetooth Low Energy mesh
- Chapter 6: Results of testing and evaluations
- Chapter 7: General conclusions and recommendations for future designs and developments
LITERATURE REVIEW
Integrated devices in sport industry
Performance tracking wearables equipped with sensors are increasingly accessible to the general public, particularly among athletes These modern technologies enable athletes and their coaches to monitor functional movements and workload effectively, ultimately enhancing performance in competitions.
Technological wearables are designed to generate real-time data that benefits both individual athletes and sports organizations Athletes utilize this information to monitor their performance and track improvements, while sports managers leverage the data from tracking devices to make informed decisions and develop effective competition strategies.
Existing commercial products
Several innovative appliances have successfully entered the commercial market, catering to the specific needs of consumers, particularly professional sports participants Notable examples of such cleverly designed and well-implemented products include the 4iii Sportiiii Sunglasses Mount, which has gained popularity among athletes.
This device integrates with various sensors worn by athletes, such as heart rate monitors and speed evaluators, to display real-time performance metrics on an LED projection Utilizing wireless communication through Bluetooth and ANT protocols, it enables athletes to monitor their health parameters effectively, enhancing their training experience.
Figure 2.2.a Sunglasses by 4iii Sportiiii [12] b, Catapult Vector
Catapult's Vector is the leading wearable device designed for field sports such as football and baseball, offering advanced features like satellite-based position tracking with precise GNSS performance Utilizing the independently developed ClearSky system, Vector ensures accuracy and reliability in real-world environments Additionally, it is equipped with a heart rate monitor and an IMU sensor, enabling athletes to effectively monitor their heart rate and running speed.
Figure 2.2.b Catapult Vector installed in sport bra [14]
The Moov HR Sweat is a cutting-edge fitness wearable designed to monitor athletes' heart rates in real time Equipped with an advanced optical heart rate sensor, it delivers highly accurate heartbeat data instantly Its innovative headband design offers greater comfort during intense workouts compared to traditional wrist devices, making it an ideal choice for serious athletes.
Figure 2.2.c Moov HR Sweat Band [16]
Technologies implemented on wearable devices
Heart rate serves as a vital indicator of physical activity, reflecting the body's physiological adaptations and exercise intensity It is essential for assessing athlete endurance by tracking VO2 changes during intense sports participation VO2, defined as the volume of oxygen consumed per minute per kilogram of body weight, has a direct correlation with heart rate; as heart rates increase, VO2 levels rise accordingly This relationship makes heart rate an accurate measure of cardio-respiratory health and fitness.
The optical heart rate sensor, commonly used in devices like wristbands and smartwatches, measures heart rate by detecting pulse waves from the athlete's wrist or fingertip This technology works by emitting infrared rays and capturing the reflected signals, allowing for accurate heart rate monitoring with each heartbeat.
+ Usable for a wide variety of sports
+ Easy and comfortable to wear
+ Affordable price for common usage
+ Low reliability, especially inaccurate in high intensity activities
Conclusion, this type of sensor is not suitable for our tracking device
Modern heart rate monitoring devices now feature advanced Electrocardiogram (ECG) sensors, which operate based on innovative technology to provide accurate heart rate measurements.
Heart pulse generates a current that is measured in millivolts using multiple electrodes placed on the body These electrodes detect the current and send the voltage readings to a sensor, which then displays the data as an electrocardiogram (ECG) on a monitor.
+ Highly accurate in heart beat tracking
+Require a good background knowledge to operate
+Easily be tangled due to many wires and leads
Therefore, ECG sensor is a suitable option for our target
Figure 2.3.1.c Holter monitor with ECG reading [21]
Body temperature is a crucial physiological metric for athletes, as fluctuations in temperature can reveal their physical endurance levels Commercial temperature sensors, equipped with telemetric core temperature technology, are integrated into wearable devices to effectively monitor this vital data.
Conclusion, this sensor is not suitable for our project
Figure 2.3.2 PouchPASS – body temperature indicating wearables [22]
A pedometer, or step-counting device, is a simple health tool that counts the number of footsteps by detecting when an internal lever arm surpasses a specific force threshold Widely used by the public, this device helps athletes monitor their energy expenditure and supports effective health and weight management.
Conclusion: this sensor is not an appropriate option for our work
Accelerometers measure the linear motion, shock, and vibration of objects by utilizing the principle of a mass on a spring When an object moves, the inertia of the mass causes the spring to either stretch or compress, enabling accurate calculations of motion.
The gyroscope sensor operates based on the principle of angular momentum conservation, resembling the characteristics of a traditional gyroscope It features a rotor, a spinning wheel mounted on a pivot, enabling rotation along three axes, known as gimbals.
Magnetometers operate on Faraday's induction laws, featuring a magnetic core encased in copper coils The magnetic field produced by the current flowing through these coils magnetizes the core, enabling precise magnetic field measurements.
The integration of accelerometers, gyroscopes, and magnetometers enables sensors to accurately track athletes' movements in all dimensions These advanced sensors are widely utilized in wearable devices, such as portable wristbands, delivering extensive physiological data that benefits both athletes and their coaches.
Conclusion, these sensors are necessary for our device
Figure 2.3.4 The Zepp system offers the type of analysis that simply wasn’t available anywhere a decade ago [27]
The Global Positioning System (GPS), also known as the Global Navigation Satellite System (GNSS), is a satellite navigation system located approximately 20,000 km above the Earth's atmosphere, providing global location and time information By utilizing trilateration, GPS-enabled devices can receive real-time location data from multiple satellites orbiting the Earth, translating this information into latitude and longitude coordinates on a map.
Figure 2.3.5.a Global Positioning System Diagram [29]
Following FIFA's approval of Electronic Performance and Tracking Systems (EPTS) in sports, GPS technology has become integral to professional athletics, particularly in football This positioning system enables athletes to monitor their movements on the field and gather vital performance data, including running speed, sprint acceleration, distance covered, and positioning By analyzing these metrics, coaching staff can design tailored training sessions and optimize competition schedules according to each athlete's physical attributes, thereby enhancing injury predictability and prevention.
Conclusion, this sensor module is suitable for our implementation
Figure 2.3.5.b Football players wear GPS tracker bras [30].
Digital signal processing
Raw data collected from sensors is often a discrete signal that contains significant noise and unwanted components, which can vary in density depending on the sampling method used Consequently, the signals from these sensors are often unreliable and impractical for real-world applications To address this issue, Digital Signal Processing (DSP) is implemented to enhance the quality of the measured signals.
Discrete signals are represented as sequences of digital number (such as 1011001)
Digital Signal Processing (DSP) comprises of many techniques to enhance the quality of signal by using some mathematic algorithms that are computed and integrated into electronic circuits [31]
Generally, Digital Signal Processing includes the following steps:
Figure 2.4.2 Block diagram of a Digital Signal Processing system
- Pre-Filter: Usually a Low-pass filter is used in this initial step to filter out the unwanted high frequency components from raw input data
An Analog to Digital Converter (ADC) transforms analog signals into digital signals, with the output data's smoothness displayed on a monitor directly proportional to the quantization resolution; higher resolution levels yield smoother output data.
Digital Signal Processing (DSP) involves the application of various functional filters tailored to meet specific problem requirements and input signals Different types of filters are utilized during this crucial step to effectively process the signals.
- DAC: Digital to Analog Converter is the module that converts the binary data back into analog signal in terms of bit-resolution
- Post-Filter: A High-pass filter is usually used in this step to remove the high frequency noise from the output signals.
Communication
The measured output signals from the sensors must be transmitted to other devices for additional processing, necessitating effective communication between them Our system utilizes two communication protocols: wired and wireless.
UART protocol is used for transferring signal externally, which means data is sent among many different circuits or modules
+ Only two wired are used to transfer data
+ Data size maximum is limited up to only 9 bits
+ Multiple Masters and Slaves are not allowed
For internal communication, SPI and I2C are used to transmit the data between devices within a circuit board
+ Multiple devices connection as Masters and Slave protocol
Our team evaluated three wireless protocols—WIFI, Zigbee, and Bluetooth Low Energy (BLE)—for our project, focusing on their physical capabilities, availability, and affordability to meet our output requirements.
Figure 2.5.2.a Bluetooth, Zigbee, and WIFI specifications comparison [33]
WIFI (Wireless Fidelity) offers the highest transmission speed and longest communication range compared to Zigbee and Bluetooth However, its significant power consumption makes it unsuitable for wearable devices that require prolonged operation.
Zigbee and Bluetooth communication protocols are crucial for our system due to their capability to establish a mesh network, enabling multiple devices to communicate with one another This feature significantly expands the range of data transfer, allowing for broader coverage and improved connectivity.
Figure 2.5.2.b Embedded Provisioner in BLE Mesh Network [34]
Table 2.5.2 Zigbee and Bluetooth comparison [35]
Emphasis Emphasizes automation Emphasizes connectivity nearby of mobile devices Power Consumption
Low power consumption, data rates on smaller packet devices
Higher power consumption, data rates on big packet devices
Frequency Around 2.4GHz 2.4GHz to 2.483Ghz
RF Channels 16 RF channels 79 RF channels
Modulation Technique BPSK, QPSK, GFSK GFSK
Coverage Covers up to 100 meters Covers up to 10 meters
Cell nodes 6500+ cell nodes 8 cell nodes
The bandwidth used by Zigbee is undoubtedly low, but it’s higher than that of Bluetooth
The bandwidth used by Bluetooth is low
In terms of specifications of Zigbee and Bluetooth protocol on Table 2.5.2, we concluded that both of which have almost the same technical characteristics that can satisfy our requirements
We chose to implement Bluetooth Low Energy (BLE) for our wireless communication protocol due to its faster data transfer rates and the affordability of its modules in local markets, enabling us to establish an efficient mesh networking system.
HARDWARE DESIGN
Technical requirements
The design of device system needs to satisfy the following input requirements:
- Speed, location, and heart rate of football athletes will be the main data targets for being calculated by this device
- Data transmission method needs to be wireless
- Extended battery capacity for continuous usage (about 3 hours)
- Outdoor and large open-space usage (about 100m length as a football stadium)
- Can take control of numerous devices and communications among them
Based on these technical prerequisites, our group can determine the way that the data is measured and communication method between devices (Figure 3.1), then selecting the most suitable electrical components
Figure 3.1 Data measurement and transmission method
Athletes' performance metrics will be gathered using specific sensor modules, including heart rate monitored by an ECG sensor, speed tracked by an IMU sensor, and position determined through GPS technology.
After receiving the necessary signal, a BLE MCU module will transfer these data wirelessly one by one in a BLE mesh, before sending these parameters to a Gateway
From there, all athletes’ information will be transmitted to a server for being calculated and stored.
Sensor specifications
The Global Navigation Satellite System (GNSS) consists of a network of space-based satellites that transmit signals to GNSS module receivers, providing essential positioning and timing data This information is crucial for accurately determining location using the GNSS system.
Global Navigation Satellite Systems (GNSS) provide comprehensive worldwide coverage, with notable examples including China's BeiDou Navigation Satellite System, Russia's GLONASS, the United States' NAVSTAR Global Positioning System (GPS), and Europe's Galileo.
- Four criteria below are used to evaluate the GNSS's working performance:
1 Accuracy: is the differences between a receiver's measured data and actual position, speed, or time
2 Integrity: is the ability of a system to provide a threshold of confidence and, in the event of a positioning data anomaly, an alarm
3 Continuity: is the capacity of a system to continue operating uninterrupted
4 Availability: The proportion of time a signal that meets the above-mentioned accuracy, integrity, and continuity requirements
To assess the device's positioning in this project, we utilize the Ublox SAM-M8Q module, which features an integrated GNSS antenna chip This enables our device to simultaneously connect to all global GNSS engines, ensuring high sensitivity and reduced acquisition times.
Main features of Ublox SAM-M8Q GNSS module:
- Having 72-chanel M8 engine, this module can communicate with many different active GNSS engines such as (GPS, SBAS, QZSS, GLONASS, Galileo)
- On-board noise filtering regulator, minimizing the surrounding noise to the lowest level
- Low power consumption based on built-in patch antenna
- 15 x 15 mm total size, facilitating it to be integrated into any designs
Table 3.2.1.2 Ublox Sam-M8Q GNSS module specifications
Navigation Update Rate Up to 18Hz
Power Consumption 29mA in Continuous mode
9.5mA in Power Save mode
A commercial Inertia Measurement Unit (IMU) used nowadays is a module which is generally a combination of the following components [39], [40]:
+ Gyroscopes: providing the angular rate measurement
+ Accelerometers: for measuring the external impacting forces
+ Magnetometers: measuring the magnetic field around the device
To achieve a three-dimensional solution, three independent inertial sensors must be combined into an orthogonal cluster, referred to as a trio, as a single inertial sensor can only measure along or around one axis.
A 3 degree-of-freedom (DOF) sensor is defined by its ability to provide a single measurement along each of the three axes (x, y, and z), typically arranged in a triad In contrast, a 6-DOF inertial system integrates both an accelerometer and a gyroscope to measure motion and orientation.
An accelerometer is a sensor that measures inertial acceleration, which refers to the rate of change of velocity over time These sensors come in various configurations, including mechanical, quartz, and MEMS types, each serving different applications in technology and engineering.
A MEMS accelerometer consists of a proof mass supported by a spring, as illustrated in Figure 3.2.2.1 The sensitivity axis indicates the direction in which the proof mass can move When linear acceleration is applied along this axis, the proof mass displaces to one side, with the degree of deflection directly proportional to the acceleration experienced.
A gyroscope is an inertial sensor that quantifies the angular rate of an object relative to its reference frame Various types of gyroscopes are available, including mechanical, fiber-optic (FOGs), ring laser (RLGs), and quartz/MEMS gyroscopes.
Quartz and MEMS gyroscopes are commonly employed in devices used by public consumers, small and medium industries, as well as fiber-optic gyroscopes
Ring laser gyroscopes generally have in-run bias stabilities (from 1°/hour to 0.001°/hour) covering tactical and navigation grades
Mechanical gyroscopes are the most common gyroscopes on the market, with in- run bias stabilities of less than 0.0001°/hour
A magnetometer is a sensor designed to measure both the magnitude and direction of a magnetic field Among the various types of magnetometers, MEMS magnetometers predominantly utilize magnetoresistance technology to effectively monitor their surrounding magnetic environment.
Magneto-resistive magnetometers utilize permalloys to change resistance in response to varying magnetic fields Commonly, MEMS magnetometers are employed to assess local magnetic fields, capturing both the Earth's magnetic field and any additional magnetic influences from nearby objects.
In this project, we utilize the MPU9255 module, which features three MEMS sensors known for their affordability, compactness, and ease of assembly in various applications While MEMS sensors may have lower accuracy compared to other types due to their susceptibility to external forces, their modest size aligns perfectly with our project objectives, making them an ideal choice for optimization.
Scale Range of Accelerometer ±2 to ±16 g
Scale Range of Gyroscope ±250 to ±2000 º/s
Scale Range of Magnetometer ±4800 àT
Sensitive Scale Factor of Accelerometer 16,384 to 2,048 LSB/g
Sensitive Scale Factor of Gyroscope 131 to 16.4 LSB/(º/s)
Sensitive Scale Factor of Magnetometer 0.6 àT / LSB
An electrocardiogram (ECG) visually represents the electrical signals associated with the physiological activities of the human heart The heart's cardiac muscles produce electricity through the contraction and relaxation of atrial and ventricular muscles, processes known as depolarization and repolarization.
ECG, displayed on paper or screen monitors as multiple waves, represents the electrical impulses generated by the heart during each beat to pump blood through the body's vessels This visual representation enables medical professionals to assess the health conditions of athletes effectively.
In this project, we utilized the DFRobot Heart Rate Monitor Sensor to detect and count heartbeats The sensor features the AD8232 main chip, which enhances signal clarity and minimizes input noise.
- Ultralow power consumption with analog-to-digital converter (170 àA average)
- DC High-pass filter and Low-pass filter integrated
- Right Led Drive (RLD) amplifier including
These above features meet the technical requirements of our design, making the AD8232 module suitable to be employed into our system
Table 3.2.3.2 DFRobot Heart Rate Sensor specifications
Operating Current < 10mA (full operation)
Integrated Hardware Filter RFI, High-pass, Low-pass
Figure 3.2.3.2 DFRobot Heart Rate Monitor sensor.
Circuit design
3.3.1 Installation position on human body
Three electrode leads (AHA) of the ECG module are attached at the chest of athletes, with the following instruction [48]:
- RA electrode (red) is placed under right clavicle (Figure 3.3.1)
- LA electrode (yellow) is placed under left clavicle, like RA electrode
- LL electrode (green) is placed under right or left abdomen to ground the electricity
Figure 3.3.1 Three leads electrode (AHA) placement on body
The system features a reliable assembly of soldered modules with supplementary functions linked to the main microcontroller, enabling the device to deliver extensive and user-friendly capabilities To meet this need, our team engineered the microcontroller's pin output circuit and performed calculations on various electronic parameters The electrical design incorporates several key components.
- Battery: Power supply for the whole circuit
- DC Converter: To maintain a constant electricity current through the system
- MCU: Undertake the main functions for the system, such as sending and receiving data from sensor modules, buttons, indicator, and making wireless communication
- GPS Module: Receive positional and speed data that was issued from the satellite and send it to the MCU for processing
- IMU Module: Measure data of angular and gravitational vector acceleration, and magnetic field, then transmit them to the microprocessor for speed and acceleration calculations
- ECG Module: Collect electrical pulse data from the heart through electrodes then send them to MCU as analog signals
- RGB Led: For multiple tasks indicating
- Power LED: To notify battery connection status
- Buzzer: For multiple tasks indicating
- Reset Button: To reset operating state of the device
- User Button: For controlling the designated functions of users
- J-Link Module: To load the program to MCU
Figure 3.3.2.1 Block diagram of the system
In order to make our device neat enough to be wearable for football athletes, we designed a 60x62cm Printed Circuit Board where all electric components are placed on
Before implementing a physical printed circuit board, a schematic sketch was constructed where all the connections of device’s components are illustrated in
The power supply unit, which is consists of 4 main components, has responsibility for providing active power to the entire system:
- A soldered micro-USB header that receives power from an LiPo battery which is the main power supply for the whole circuit
- Module TP4046 LiPo battery charging is connected to the battery to recharge it when the device runs out of power
- The HT016 module is used to boost the output voltage of battery up to 5V on demand of ECG, IMU, and GPS sensor modules
- Finally, the AMS1117 IC is implemented to dropout and stabilize the 3.3V output, which is used for other electric components
A LiPo battery serves as the power source for the system, and to ensure a minimum continuous operation of at least two hours, we calculated the maximum power consumption of the device's components to determine the appropriate battery capacity.
The maximum operating currents for various modules are as follows: the GPS module operates at 29mA, the IMU module at 6mA, the ECG module at 10mA, the MCU module at 27mA, and other electrical components at 21mA.
So, the approximate power consumption of total system is 87mA
Based on the battery life calculation [49], we get the formular:
𝐿𝑜𝑎𝑑 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 (𝑚𝐴) (1) Our device is expected to operate in at least 3 hours So:
We opted for a LiPo battery with a capacity exceeding 279mAh for our device, ultimately selecting a 1000mAh single cell LiPo battery due to its market availability.
Figure 3.3.2.2.b Single Cell LiPo Battery 1000mAh
The TP4056 module employs a constant current and constant voltage (CC-CV) charging mechanism specifically designed for single cell lithium-ion batteries Upon connection, the chip detects the battery voltage and delivers a regulated charging current, which can be adjusted using a 1kΩ resistor connected to the IN pad Additionally, the module features overcharge and deep discharge protection, ensuring safety during the charging process Once the battery reaches full charge, the module automatically halts the charging process.
Table 3.3.2.2.b TP4056 charging module specifications [51]
To meet the 5V input voltage requirement of all sensor modules in this device, we utilized an HT016 module, which boosts the output voltage by adjusting its resistance with a manual potentiometer.
The output current of TP4056 is fed to the input pin of HT016 module Then the output current of HT016 would be used for the whole circuit
Table 3.3.2.2.c HT016 voltage booster module specifications [52]
Figure 3.3.2.2.d HT016 Voltage Booster module
The AMS1117 is a linear voltage regulator designed to adjust voltage from input to output, providing a stable output voltage of 3.3V Capable of delivering up to 1A of current, this integrated circuit is essential for voltage stabilization in various electronic applications.
Figure 3.3.2.2.e AMS1117 IC b, Indicator Block:
This block contains 1 red LED to indicate whether power is on or not, 1 programable RGB LED for other designations, and 1 buzzer for functional alarming
This block contains 1 Reset Button to reset the device’s parameters to initial state, and 1 User Button to enable and disable the connection between many devices and the computational gateway
Figure 3.3.2.2.h Buttons d, Sensor Modules Block:
The sensor module block utilizes three specified sensor modules to collect measured parameters, which are then transmitted to the MCU for processing calculations The MCU also has the capability to write data back to the sensor modules, generating the desired data stream Connections from the MCU to the sensors can be made via pins or direct soldering to the board.
Figure 3.3.2.2.i Sensor Module Block schematic
Figure 3.3.2.2.j MCU Module Block schematic
We utilized the E73-2G4M08S1C as our primary MCU module, which features the nRF52840, a compact and energy-efficient wireless Bluetooth module developed by Ebyte This module supports BLE 4.2 and 5.0, powered by the robust ARM Cortex-M4F core and equipped with various peripherals including UART, I2C, SPI, ADC, and DMA.
- The measured communication distance up to 120m
- Maximum transmission power of 6mW, ultra-low power consumption
- Built-in ceramic antenna, no external antenna required
Bluetooth Working Frequency 2.4GHz (32MHz oscillator)
3.3.3 Printed Circuit Board (PCB) design
Circuit board dimension: According on the selected electrical components, the back cover has a 60mm width and 62mm height
We utilize Altium Designer for creating and refining schematics, whether they are simple or complex The process of verifying electrical regulations becomes straightforward when using components from the official Altium library or by incorporating new custom symbols.
With Altium's 3D Viewer capability, we can display realistic renderings of our design to verify for mechanical fit or to preview the final product
The scopes of us are to arrange all the Electrical components neatly so that they will not be taking up much space and are simple to route:
- The power supply block, which generates noise, needs to be placed away from the NRF52840 Bluetooth antenna, IMU module, and I2C communication lines
- We designed pins for simple module modification or plugging to link the board and other modules
- In order for the system to run without any voltage drop that would cause system instability, the booster circuit is integrated
- To prevent interfering with other circuit modules, it will also be segregated from the battery
Figure 3.3.3.b Electrical Components arrangement on the PCB
After assembling all components, we initiate the routing process according to the PCB design, carefully adjusting wire angles and incorporating vias to ensure they can navigate the board's backside without contact The red and green lines indicate the top and bottom layers of the circuit board, respectively.
The following rules are used for soldering part:
- Minimum clearance between two wires is greater than the width of the wire
- Width: The wire's width varies depending on whether it is a control signal line, a 5V DC line, a 12V DC power line, or a line that outputs four coils
- Routing via: the via itself, the copper plates' size, shape, and dimensions
- Silk to Silk Clearance: This term refers to the minimal distance between components, which is based on their individual sizes
- Mask for Silk Solder Setting the bare minimal distance between components and the solder mask is known as clearance
Figure 3.3.3.c Top and bottom layer of PCB
Figure 3.3.3.d Top and bottom view of mainboard
Table 3.3.3 Electrical components on one PCB
CONTROL SOFTWARE DEVELOPMENT
Data processing
The input data often contains significant noise and unwanted elements Considering our expertise, available technology, time constraints, and budget limitations, our team has chosen to integrate specific techniques into our device system These methods are designed to process raw data inputs, improving output quality and minimizing noise effectively.
Our focus will not be on in-depth data analysis, as it falls outside our area of expertise Consequently, several unresolved issues and challenges related to sensor data processing will remain We encourage other data specialists to take the lead in addressing these developments.
GPS data processing
4.2.1 Data transmission of GNSS system
GNSS satellites operate in precise orbits, providing real-time signals about time and position to Earth twice daily GNSS receivers capture these signals and decode them through trigonometric calculations, enabling accurate navigation.
Smartphones and GPS tracking modules require signals from a minimum of three satellites to accurately deliver two-way location and motion tracking information, including travel speed, direction, and journey distance.
Data information received from satellites is used globally, so it needs to be in an international form This form is called NMEA
The National Association of Marine Electronics (NMEA) establishes the standard packet format for most GNSS receivers currently in use This standard format organizes various information elements, which are separated by commas, within GNSS packets.
Figure 4.2.2.a Example of GNSS sentence format [58]
String Identifier structure: Including 5 characters
- The first 2 characters defined the GNSS systems that data is received from:
+ GB/BD: BeiDou satellites only
+ GN: Satellites from multiple systems
- The last 3 characters defined the data type, the most common used formats:
+ GLL: Latitude and Longitude data
+ VTG: Vector track speed over the ground
+ GGA: Provides 3D location and accuracy data
Figure 4.2.2.b GNSS data received from our device
Figure 4.2.2.b above illustrates some GPS data that received from our device
- GNGLL: Latitude and Longitude information that was received from multiple GNSS systems
In this project, we focused on collecting athletes' positional data, utilizing latitude, longitude, and time for accurate tracking Consequently, our system exclusively employed the GLL format to achieve this objective.
Using the Google Map API and Python, we created two rectangular walking paths based on satellite position data collected during our device testing on the HCMUTE football field, as shown in Figures 4.2.3.a and 4.2.3.b The illustrations accurately depict the device's real location, although there are some discrepancies of approximately ±2.5 meters compared to the actual position.
Figure 4.2.3.a First rectangular path sketched from GPS data
Figure 4.2.3.b Second rectangular path sketched from GPS data
Using data from a GPS module, we create a Heat Map to visually represent the frequency of our device's locations on the HCMUTE football field This technique employs color to indicate the magnitude of occurrences, utilizing Python and the Google Maps API to illustrate the latitude and longitude data effectively.
Figure 4.2.4 Heat Map illustration of our device in HCMUTE football field
Figure 4.2.4 shows the Heat Map result on the football field The darker color in the map (red) means the higher appearing frequency of device at these positions
Football experts can analyze statistical data on athletes' performance during games, enabling them to create tailored training plans and exercises for each individual player.
ECG data processing
An electrocardiogram (ECG) captures the electrical signals of the heart, reflecting voltage fluctuations from action potentials in excitable cardiac cells Electrodes measure these fluctuations to provide insights into heart activity By analyzing ECG data, healthcare professionals can diagnose health issues and assess the fitness levels of athletes.
An ECG waveform from a healthy individual consists of three key segments: the P wave, representing atrial depolarization; the QRS complex, which reflects both atrial repolarization and ventricular depolarization; and the T wave, indicating ventricular repolarization An ideal ECG signal is depicted in Figure 4.3.1.c.
Figure 4.3.1.a Anatomy of the heart and its electrical conduction system [59]
Figure 4.3.1.b An ECG cycle with waves and segments [60]
Figure 4.3.1.c Theoretically ideal ECG signal
ECG signals, like other forms of analog data, are susceptible to various unwanted noise from electric devices, the surrounding environment, and the human body, which can adversely impact measurement accuracy The three main types of noise affecting ECG readings are [61]:
1 Baseline wander (BW) noise: BW is known as low frequency noise (between 0.05 and 2Hz) which is caused by human respiration, body movements, low quality electrodes, and impedance of skin area where electrodes are attached This will result bad in the ECG data output by varying its horizontal line up and down, making the signals unreliable and be harder for heart beat detecting procedure
2 Power – line interference noise: This noise is caused by the power suppliers such as batteries or other surrounding electrical devices, which have 50 or 60Hz
71 frequency Because of the resonance of 50/60Hz frequency from these devices with high frequency in ECG signals, the baseline would be thickened
Figure 4.3.1.e 50Hz frequency interference noise in ECG
3 Muscle artefacts (EMG noise): Electromyography (EMG) noise is caused by electricity from muscles movements which have frequency range varying from 20 to 150Hz Most of the time EMG noise will originate from upper part of body, such as eyes blink, neck moving, raising arms, swallowing, etc This affects severely bad on ECG signals making chaotic distortion in the results
Figure 4.3.1.f EMG noise in ECG
To filter out the baseline wander noise from ECG signals, we employed a Butterworth High-pass filter, which can cancel out the low frequency noise of the signals
The Butterworth filter is known for its nearly ripple-free output response in both the pass band and stop band, making it a maximally flat filter compared to other digital filters As the order of the filter increases, the response becomes even flatter; however, designing filters of excessively high order can lead to distortion in the results.
The transferred function of nth-order Butterworth high-pass filter [63]:
We used MATLAB as our platform for designing digital filters MATLAB is a reliable platform for engineers and scientists to analyze many different related problems
Figure 4.3.2.c MATLAB filter design tool
A Butterworth High-pass filter is generated by editing these following parameters:
After adjusting necessary parameters, we imported a raw ECG signal that was measured from our body by ECG sensor
As seen in Figure 4.3.2.d, the raw signal contains a lot of noise, such as baseline wander, powerline interference, and EMG noise Dramatically signal fluctuations caused our detection more tricky
Figure 4.3.2.d ECG raw signal that contains many types of noise
After Butterworth high-pass filter applying, the low frequency noise (baseline wander) was almost attenuated and disappeared, especially in the half later samples, resulting in a much cleaner signal (Figure 4.3.2.e)
Figure 4.3.2.e ECG signal after Butterworth high-pass filter was applied
We effectively eliminated low-frequency noise and proceeded to cancel high-frequency interference, including powerline noise at 50/60Hz and muscle noise ranging from 30 to 150Hz To achieve this, we utilized a filter design tool to create a Butterworth low-pass filter.
- Frequency Specifications: Fs = 200Hz, Fc = 25Hz
Figure 4.3.2.f ECG signal after Butterworth low-pass filter was applied
After eliminating high-frequency noise, the signal data appears significantly cleaner, as illustrated in Figure 4.3.2.f The integrity of the cardiac cycle segments, particularly the R waves essential for heart rate (BPM) detection in subsequent analyses, was nearly flawlessly maintained Our effective processing techniques ensured the collection of the clearest signals possible.
Heartbeat detection is a crucial step in measuring the heart's pulse, which is essential for calculating a person's heart rate in beats per minute (BPM).
This step is based on a peak detection technique that was developed by Pan- Tompkins, so it is called Pan-Tompkins’ algorithm for heart beat detection
Figure 4.3.3.a Pan-Tompkins’ algorithm procedure [64]
Low Pass Filter and High Pass Filter were completely implemented so two first steps will be ignored, continuing to employed other stages left
The Differentiation or Derivative step is essential for emphasizing the slope of the QRS complex, which features a significantly steeper slope compared to other waves in the cardiac cycle This distinct characteristic allows for the identification of these segments as the highest peaks, facilitating more accurate analysis.
Squaring technique in the next step is employed in order to amplify the amplitude of the peaks, which were found before, to the power of 2, making them non-negative
After that, a Moving Windows Integration technique is used to smooth the signals by canceling out their fluctuations
To identify the R peaks, the primary focus of our analysis, we applied a carefully calibrated threshold to the output from the previous step This threshold was meticulously adjusted to ensure that all values exceeding it were included, while lower data points were effectively excluded.
By implementing this procedure in MATLAB, we can accurately detect all R peaks in the ECG signal and calculate the heart rate As illustrated in Figure 4.3.3.b, the R peaks are highlighted in red, resulting in a calculated heart rate of 96 BPM.
Medical specialists can assess an athlete's health based on their heart rate With a heart rate of 96 BPM and normal cardiac cycle segments, it can be concluded that this individual is in good health and fitness.
Figure 4.3.3.b R peaks detection and heart rate computed
IMU data processing
The MPU [65] module operates with a default sampling rate of 1 kHz, which can be programmatically adjusted Utilizing MEMS sensors, the module is highly sensitive to noise and value deviations caused by vibrations or external forces This is particularly true for magnetometers, which can be influenced by the Earth's magnetic field and electrical noise from nearby magnetic components like batteries or power lines To ensure the most accurate output data, regular calibrations are essential.
The calibration of the accelerometer and gyroscope utilizes the DMP library from InvenSense Although offset values are pre-stored in the sensor processor chip's memory, various environmental factors in different active regions necessitate readjustment and recomputation of these values in the registers to enhance sensor accuracy.
The working principle of offset registers involves applying shifts to all data from MEMS sensors prior to transmitting this information to the registers for user access This offset application occurs before any FIFO and DMP processing, ensuring that the data stored in the FIFO, output registers, and DMP already incorporates these offsets.
Figure 4.4.2.a Data Signal Processing Diagram of Gyro and Accel
Biases for accelerometers and gyroscopes can be established through various methods, fundamentally relying on the principle that the device is positioned in a known orientation, with the MPU components oriented either upwards or downwards.
The stationary device should ideally produce a gyro output of [0, 0, 0] and an acceleration output of [0, 0, +1G] By collecting samples from each axis, we can compute the average offset, known as the Offset Biases, based on these ideal values.
Figure 4.4.2.b Raw data of accelerometer
Figure 4.4.2.c Raw data of gyroscope
Figure 4.4.2.d Calibrated data of accelerometer using DMP library
Figure 4.4.2.e Calibrated data of gyroscope using DMP library
An ideal three-axis magnetometer measures magnetic field strength along the orthogonal X, Y, and Z axes, capturing the Earth's magnetic field without interference As the sensor rotates through all orientations on a sphere, the recorded data reflects the magnetic field's strength, represented by the radius of this sphere Calibration concepts, including hard-iron and soft-iron effects, are essential for accurate measurements.
Noise sources and manufacturing defects can significantly impact magnetometer measurements, with hard iron interference being particularly prominent These hard iron effects introduce stationary magnetic noise, often resulting from nearby metal objects on the same circuit board Consequently, hard iron disturbances alter the reference point of the ideal magnetic field.
On the other hand, the effects caused by soft iron are more modest They are originated from the elements which are close to the sensor, modifying the magnetic field
This causes the ball to stretch and tilt with ideal dimensions As a result, the obtained measurements are ellipsoidal
To replicate the effects of a soft iron magnetic field, we initially rotated the geomagnetic field vectors from the IMU to the sensor frame, followed by a stretching process, and concluded with a rotation to the global frame This calibration technique is essential for accurate magnetic field measurements.
The magnetometer calibration is computed with this equation:
- 𝑥: is raw value vector with 1x3 dimensions
- 𝑏: is vector 1x3 dimensions for hard-iron offsets
- 𝐴: is 3x3 matrix for soft-iron raw values
To calibrate the magnetometer, which includes vector "𝑏" and matrix "𝐴", we first established the coordinate position We then rotated the sensor in a spherical orbit to sample the magnetometer, ensuring we gathered the most accurate parameters for our calculations Once the sampling process was complete, we incorporated the magnetic norm and the sampled parameters into our calculation tool to derive the two parameter vectors.
The value of matrix A and the value of vector b defined by using MATLAB are:
We established the necessary values for soft-iron and hard-iron calibration, applied these values to calibrate the magnetometer, and then collected and compared the raw and calibrated data Finally, we graphically represented the results to validate our findings.
Figure 4.4.3.a illustrates that after rectification, the center of the corrected data aligns with the zero origin, while the raw data's center is displaced from this point This suggests that the rectification process was successful.
Figure 4.4.3.a Comparison of raw data and calibrated data in x-y axis
Figure 4.4.3.b Comparison of raw data and calibrated data in y-z axis
Figure 4.4.3.c Comparison of raw data and calibrated data in x-z axis
The acceleration values obtained from the Inertial Measurement Unit (IMU) are vector quantities along the x, y, and z axes To determine the athlete's true acceleration during movements, it is essential to convert these vector values into global acceleration measurements.
The orientation of an Inertial Navigation System (INS) relative to the global reference frame is monitored using angular velocity signals, as described by the following equation.
To define the orientations of an Inertial Navigation System (INS), various situational representations can be employed, including Euler angles, quaternions, and direction cosines This section focuses on the directional cosine representation to develop an attitude tracking algorithm, while also incorporating similar derivations found in Euler angles and quaternions.
The attitude of the body frame with respect to the global frame is given in direction cosine format by a [3x3] C rotation matrix, where each column is a unit vector along one
85 of the body axes Global axes are specified In the body frame, a vector quantity 𝒗 𝑏 equals the vector:
𝒗 𝑔 = 𝑪𝒗 𝑏 (5) When the global frame is defined The inverse transformation is represented by:
To compute the rotation matrix, we start by utilizing the AHRS-Madgwick update to obtain the quaternion value, which serves as the basis for deriving the C rotation matrix.
Sebastian Madgwick presented this orientation filter for IMUs with 3-axis gyroscopes and accelerometers and MARG arrays with 3-axis magnetometers
WIRELESS CONNECTION OF DEVICES
Connection Description
Figure 5.1 Communication diagram of the system
Connections among modules within the device are physical connections between sensor pins and processor pins using TWI, UART as wired communication
The BLE protocol enables wireless data transmission between devices, allowing athletes in proximity to send data directly to the client This data is subsequently relayed to the gateway through UART communication.
Requirements of devices
Based on the realistic demands technical specifications of devices in the system, we concluded the most significant requirements for implementation:
- Connection Interval: this parameter will determine the frequency of data transmission between wireless devices A maximum and minimum transmission interval
90 will be supplied when an update is requested from any peripherals The most efficient interval is between 7.5ms and 4s
To achieve rapid data transmission between nodes in the mesh network, we have implemented zero latency settings However, this configuration keeps the peripherals in a sleepless mode, resulting in increased power consumption for the system.
Each message in the data payload includes three key pieces of information: position, speed, and heart rate The position data, which is 8 bytes, is transmitted every 0.1 seconds, while the speed data, consisting of 2 bytes, is also sent every 0.1 seconds Additionally, heart rate information is transmitted as 4 bytes once per second.
The system's low power consumption necessitates that each device transmit data within a maximum distance of 10 meters from one another, resulting in reduced connection times and conserving power resources for each module.
Wireless Communication Topology
Wireless connections utilizing the Bluetooth standard encompass point-to-point, broadcast, and mesh types However, the broadcast method, which transmits data without confirmation of receipt by other devices, is not appropriate for our project and will not be implemented.
The term topology is commonly used in the networking field The topology of a network can be defined either physically or intellectually
Logical topology defines the data flow within a network, while physical topology describes the arrangement of nodes A direct connection between two nodes through a wire or other medium is known as point-to-point topology.
A point-to-point topology can be exemplified by the relationship between a remote control and a television at home In this setup, a direct connection is established when the user operates the television using the remote, illustrating the concept of point-to-point communication.
The Bluetooth Low Energy standard operates on a point-to-point star architecture, which restricts its network range In this star topology, the master node serves as the central hub, while the surrounding slave nodes rely on it for data transfer, as they cannot communicate directly with one another.
Table 5.3.1.1 Point-to-point BLE specifications [74]
Optimized Feature Short burst data transmission
Max connections/device (piconet) Unlimited (implementation specific)
Data rate 125 Kb/s to 2 Mb/s
Security 128-bit AES, user defined application layer
A mesh network consists of interconnected nodes that communicate with each other To join this network, a device must undergo a five-step provisioning process, which is essential for ensuring security.
1 Beaconing: Device transmits advertising packets to other nearby devices
2 Invitation: One node transmits a provisioning invitation PDU others
3 Public key exchange: Party exchange public keys for encryption and decryption
4 Authentication: The provisioner verifies the connecting node using a proper authenticated method
5 Provisioning data distribution: Provisioner generates and sends provisioning data to the verified node
Once a device is provisioned, it gains the ability to send and receive messages within the mesh network However, certain functionalities—specifically relaying, proxying, friend, and low power—are limited to nodes equipped with these specific features.
Figure 5.3.1.2 Example connection of BLE mesh [76]
- Elements in our BLE mesh networking:
Relay nodes facilitate communication between distant network devices by receiving and retransmitting messages until they reach their intended destination While they enable messages to traverse the entire network, the effective distance a message can travel is constrained by the time-to-live variable, which is essential for optimal network management.
Proxy nodes in Bluetooth mesh networks, introduced in 2017, enable older Bluetooth devices that lack the BLE mesh stack to connect through the GATT interface This capability allows these devices to exchange data with the mesh network using GATT's read and write operations.
Low Power and Friend Nodes work together to optimize network efficiency Nodes with limited energy resources can activate a low power mode, allowing them to send messages at scheduled intervals Instead of transmitting all data, these nodes receive necessary information as needed, ensuring effective power management.
94 the buddy node, which transmits the data to the low power node when the friend node demands it A "friendship" relationship exists between the two nodes
Up to 16384 group addresses Supports publish/subscribe addressing
Max payload size 29 – byte payload
Device, network, and application levels
Service definition Mesh models, mesh properties
5.3.2 BLE mesh setting up procedure
Our objective is to develop a method to publish sensor data from athletes to a client and send all that information to our PC to visualize their performance
We enhanced the Generic on/off model standardized by Bluetooth SIG to accommodate larger message payloads The client's model remained unchanged, as it solely receives data; therefore, we eliminated the limitations on the packet length that the client can accept.
The server was significantly changed by our modification in order for the model to meet our desirable requirements
Furthermore, we took advantages of an existing status message to send our data, made some tweaks to transfer bigger messages, and enlarged the data payload of the transferred messages
To enhance transmission speed and reduce the risk of data loss associated with large, unacknowledged messages, we ensure that the transferred data consists of byte messages rather than string types.
Our mesh contains 3 devices that act as servers containing sensor data and 1 client waiting for data before transmitting them to the Gateway The network could be visualized in Figure 5.3.2.a
After flashing the firmware to all devices, each of them will act as provisioners needing to be added into a mesh networking and behaving as a member of the mesh
Nordic Semiconductor provides us an app that can speed up the development stage called nRF MESH It helps us to test our software without spending time on creating a provisioner
Once the mesh is established, server model instances can exclusively publish to the client and subscribe to all existing nodes, while client model instances have the capability to publish to all types of nodes.
After completing the provision and configuration steps, our system operates as server instances that gather data before transmitting it to the client Once the sensor information is received, the client can then send the data to the PC This process is summarized in Figures 5.3.2.b and 5.3.2.c.
Figure 5.3.2.b Server Instance flow chart
Figure 5.3.2.c Client Instance flow chart
RESULTS AND EVALUATIONS
Hardware result
Our PCB components functioned flawlessly, with all sensor modules successfully receiving and transmitting their designated data signals to the MCU The main MCU module, equipped with integrated BLE protocol, established efficient mesh networking communication among devices Additionally, all other electrical components operated without any errors.
Data processing results
GPS data transmission is highly accurate, but in urban areas with tall buildings, satellite signals can be obstructed and disrupted Consequently, the optimal environment for our device to function effectively is in open spaces like fields or large stadiums.
After having acceleration data, we can determine the velocity of athletes by using mathematic integration method, the result is illustrated in Figure 6.2.2
As being shown, the maximum speed of us is about 2m/s, which is 7km/h This means we were walking at the testing time
To measure electrical signals from the heart, we utilized the Better Serial Plotter app, which allowed us to receive data via a UART port and display it in real-time on our laptop screen.
The raw ECG signal depicted in Figure 6.2.3.a illustrates the presence of significant unwanted noise, including baseline wander, power-line interference, and EMG noise, which can affect the accuracy of the readings.
To eliminate baseline wander noise, we utilized a 3rd order high-pass Butterworth filter with a 2Hz cutoff frequency, effectively reducing low-frequency components As shown in Figure 6.2.3.b, this filtering process successfully removed baseline wander noise, leading to a significantly more stable signal without chaotic fluctuations.
Figure 6.2.3.b ECG signal after high-pass filter being applied
The ECG signal is often affected by high-frequency components, including power-line interference and EMG noise, which can distort its appearance To address these issues, we implemented a 15th order low-pass Butterworth filter with a cutoff frequency of 30Hz in our system The effectiveness of this filtering process is illustrated in Figure 6.2.3.c below.
Figure 6.2.3.c ECG signal after low-pass filter being applied
The signal has been significantly improved, resulting in a cleaner and clearer output High-frequency components have been effectively attenuated and removed, while all segments and waves of the cardiac cycle have been perfectly preserved, paving the way for the next processing stage.
Using MATLAB as a programming platform, we applied the Pan-Tompkins technique to detect all R peaks in the signal, resulting in a calculated heart rate of 91.2 BPM Figure 6.2.3.d illustrates the detected R peaks, highlighted in red, confirming the heart rate measurement.
Figure 6.2.3.d Heart beat detection by MATLAB
The final calculation indicates that a heart rate of 91.2 BPM falls within the warming-up stage for athletes Additionally, all peaks, including the R peak and T wave, are normal, suggesting that the individual's health condition is in good standing.
Connectivity Results
We have successfully established a mesh network with three active devices, transmitting data back to the client, as illustrated in Figure 6.3 However, after a period of operation, the system experienced message loss, likely due to the use of unacknowledged messages, which do not ensure consistent delivery to the client.
By integrating a GPS sensor, we can analyze the connectivity of our device As illustrated in Figure 6.3.b, a testing scenario was established using three BLE devices to form a mesh network, enabling the reception of position data from the GPS sensor.
The initial design of the red and blue paths was meant to be rectangular and triangular; however, data loss during transmission led to distorted and uneven shapes on the football field.
Figure 6.3.b BLE connection testing using GPS data
CONCLUSION AND RECOMMENDATION
Conclusion
In conclusion, our GPS module performed effectively in open spaces with minimal obstructions, demonstrating reliable data signal reception While some positional inaccuracies were noted, the system is adequately suited for use in football fields with few obstacles, making it a practical solution for real-life applications.
ECG conclusion: the device can receive signal from human body and make some mathematic calculations for observing a better-quality result Therefore, this module is suitable enough to be employed
In conclusion, while speed parameters can be derived from the IMU sensor, the presence of significant errors renders this data unreliable Therefore, it is essential to conduct a thorough examination of the sensor before considering its use in future applications.
In conclusion, while our device has shown notable advancements in data transfer using a mesh protocol, it is important to note that this technology is still in the testing and development phases and is not yet suitable for real-world applications.
In general conclusion, our device system cannot satisfy the professional usage in football Therefore, it needs to be developed further in the future.
Recommendations and future developments
Future developments of our system should focus on enhancing hardware components, including a protective housing that safeguards internal elements and ensures a sleek design suitable for athletes This improvement is essential for creating a wearable device comparable to the Catapult Vector, as illustrated in Figure 7.2.a below.
To ensure its practical application in real-life competitions, the device should be designed to attach securely to athletes' bodies using holding mediums like sports bras or wearable belts during football matches.
Figure 7.2.b Adidas Football Sport Bra
[1] “The Global Sports Technology Market size is expected to reach $48.7 billion by
2028, rising at a market growth of 16.8% CAGR during the forecast period” Author: ReportLinker September 29, 2022
[2] “History and FIFA Regulations of EPTS” Author: Noor Tasnim, December 19, 2022
[3] “Approval of Electronic Performance and Tracking System (EPTS) devices” Author: FIFA July 8, 2015
[4] “Manchester City launches a new wearable performance tracker” Author: Victoria Song November 4, 2022
[5] “Newcastle United Turns to Wearable Technology for Safe Premier League Return” Author: Robert Kidd June 16, 2020
[6] “Premier League Club Fulham FC Expands Playermaker Wearable to Academy” Author: Andrew Cohen May 28, 2022
[7] “Manchester City launches a new wearable performance tracker” Author: Victoria Song November 4, 2022
[8] “Giải mã chiếc áo công nghệ đặc biệt góp công lớn giúp đội tuyển Việt Nam thắng Indonesia” Author: Anh Việt June, 2019
[9] “Áo đồng phục gắn chip GPS của đội tuyển bóng đá Việt Nam” June, 2019
[10] “Catapult Debuts New GPS Wearable for Amateur Athletes” Author: Joe Lemire August 18, 2021
[11] “In Depth Review of the 4iiii Sport-iiiis ANT + Heads Up Display System” Author:
[12] “In Depth Review of the 4iiii Sport-iiiis ANT + Heads Up Display System” Author:
[13] “10 REASON WHY CATAPULT VECTOR IS THE MOST ADVANCED ATHLETE TRACKING TECHNOLOGY IN SPORT” Author: catapultsports.com March 14, 2019
[14] “Fitness Vest Helps Amateur Athletes Track & Train Like a Pro” Author: JOSEPH SCOTTING August 18, 2021
[15] “Moov HR Sweat review” Author: Michael Sawh July 3, 2017
[16] “10 Most Innovative Wearables for Fitness and Sport” Author: APAC BUSINESS headlines
[17] “Relationship Between Heart Rate & VO2 Max” Author: Emily Alvers MS, CSN, CPT
[18] “Muscle Oxygen Training” Author: BJRMD September 20, 2019
[19] “Optical Heart Rate Monitoring: What you need to know” Author: pinnacle LIFESCIENCE
[20] “Wearable Performance Devices in Sports Medicine” Author: James E Voos, MD, Department of Orthopaedic Surgery, University Hospitals Case Medical Center, 11100 Euclid Avenue, Cleveland, OH 44106 January, 2016
[21] “Holter Monitor” Author: JOHNS HOPKINS MEDICINE
[22] “This wearable constantly does a temperature-screening of your body so you don’t have to” Author: SARANG SHETH August 31, 2020
[23] “Pedometer for Weight Loss – How a Pedometer can help you to Lose Weight”
[24] “MEMS Accelerometers” Author: SILICON SENSING
[25] “What is a Gyroscope Sensor: Working & Its Applications” Author: WatElectronics May 16, 2020
[26] “An introduction to magnetometers, where they are used and how they work” Author: Team YoungWonks February 16, 2021
[27] “Zepp Labs outs 3D swing analysis systems for baseball, golf and tennis”, Author: Mike Hanlon November 14, 2013
[28] “Principles of GPS” Author: Carl Carter February, 1997
[29] “Global Positioning System and GPS Devices” Author: Rashid Faridi August 8,
[30] “GPS Tracking Creates More Competitive Players” Author: SPT
[31] “An Introduction to Digital Signal Processing” Author: Steve Arar June 30, 2021
[32] “Wired Communication Protocols in IoT” Author: Aparna Chaurasia
[33] “Differences between Bluetooth and Zigbee” Author: MKS075 July 12, 2022
[34] “Embedded Provisioner for BLE Mesh Network” Author: Michal October 20, 2021
[35] “Difference Between Zigbee and Bluetooth” Author: Sandeep Bhandari
[36] “What is GNSS” Author: EUSPA December 3, 2021
[37] “SAM-M8Q module Product summary” Author: U-Blox
[38] “GPS MODULE SAM-M8Q” Available: mateksys.com
[39] “WHAT IS AN INERTIAL MEASUREMENT UNIT” Author: vectornav.com
[40] “MEMS OPERATION” Author: vectornav.com
[41] “TDK InvenSense MPU-9255” Available: utmel.com
[42] “MPU-9255 Product Specification” Author: InvenSense Inc September 14, 2014
[43] “Hướng dẫn sử dụng module cảm biến gia tốc góc – MPU9255” Author: mlab.vn
[44] “Basic Principles of Electrocardiography” Author: Johnson Francis, MBBS, MD,
DM, Former Professor of Cardiology, Calicut Govt, Medical Kozhikode, Kerala, India December 27, 2022
[45] “Electrocardiogram (ECG or EKG)” Author: American Heart Association (AHA) December 7, 2022
[48] “ECG Lead positioning” Author: Mike Cadogan January 30, 2022
[49] “Batter Life Calculator” Source: digikey.com
[52] Mạch tăng áp DC-DC HT016 Source: hshop.vn
[54] E73-2G4M08S1C User manual v1.9 Source: ebyte.com
[56] “Global Navigation Satellite Systems” Author: Hemisphere January 23, 2022
[57] “What exactly is GPS NMEA data?” Author: Eric Gakstatter February 4, 2015
[58] “Read data from GPS receiver” Source: mathworks.com
[59] “The Electrocardiogram – A Brief Background” Author: Leif Sửrnmo and Pablo Laguna December 2005
[60] “An ECG T-wave Anomalies Detection Using a Lightweight Classification Model for Wireless Body Sensors” Author: Medina Hadjem and Farid Nạt-Abdesselam September 2, 2015
[61] “Review of noise removal techniques in ECG signals” Author: Shubhojeet Chatterjee, Rini Smita Thakur, Ram Narayan Yadav, Lalita Gupta, Deepak Kumar Raghuvanshi December 01, 2020
[62] “Butterworth Filter Design” Source: electronics-tutorials.ws
[63] “Butterworth Filters” Author: Pieter P July 15, 2021
[64] “Swarm Intelligence Approach to QRS Detection” Author: Mohamed Belkadi and Abdelhamid Daamouche July 4, 2020
[65] “MPU Hardware Offset Registers Application Note” Source: InvenSense
[66] “How to Calibrate a Magnetometer” Author: ShawnHymel
[67] “Magnetometer Calibration” Source: mathworks.com
[68] “Madgwick Orientation Filter” Author: Mario Garcia 2019
[69] “Bluetooth Wireless Technology” Source: bluetooth.com
[70] “Resources for Bluetooth Low Energy” Author: Junqing Zhang November 7, 2020
[71] “Point to Point Topology” Author: Pooja Gupta
[72] “A Bluetooth Low Energy real-time protocol for Industrial Wireless mesh Networks” Author: G Patti, L Leonardi and L Lo Bello 2016
[73] “SELF-ORGANIZING ROUTING PROTOCOL FOR BLUETOOTH LOW ENERGY SENSOR NETWORKS” Author: Marko Haatanen 2012
[74] “Devices need multiple ways to connect” Source: Bluetooth.com
[75] “An auto-configuring mesh protocol with proactive source routing for Bluetooth Low Energy” Author: Julio León 2018
[76] “A survey on Bluetooth multi-hop networks” Author: Nicole Todtenberg and Rolf Kraemer June 8, 2019