Ensuring athlete physical fitness using Cyber-Physical Systems (CPS) in training environments

Abstract

BACKGROUND:

Sports have been a fundamental component of any culture and legacy for centuries. Athletes are widely regarded as a source of national pride, and their physical well-being is deemed to be of paramount significance. The attainment of optimal performance and injury prevention in athletes is contingent upon physical fitness. Technology integration has implemented Cyber-Physical Systems (CPS) to augment the athletic training milieu.

OBJECTIVE:

The present study introduces an approach for assessing athlete physical fitness in training environments: the Internet of Things (IoT) and CPS-based Physical Fitness Evaluation Method (IoT-CPS-PFEM).

METHODS:

The IoT-CPS-PFEM employs a range of IoT-connected sensors and devices to observe and assess the physical fitness of athletes. The proposed methodology gathers information on diverse fitness parameters, including heart rate, body temperature, and oxygen saturation. It employs machine learning algorithms to scrutinize and furnish feedback on the athlete’s physical fitness status.

RESULTS:

The simulation findings illustrate the efficacy of the proposed IoT-CPS-PFEM in identifying the physical fitness levels of athletes, with an average precision of 93%. The method under consideration aims to tackle the existing obstacles of conventional physical fitness assessment techniques, including imprecisions, time lags, and manual data-gathering requirements. The approach of IoT-CPS-PFEM provides the benefits of real-time monitoring, precision, and automation, thereby enhancing an athlete’s physical fitness and overall performance to a considerable extent.

CONCLUSION:

The research findings suggest that the implementation of IoT-CPS-PFEM can significantly impact the physical fitness of athletes and enhance the performance of the Indian sports industry in global competitions.

Keywords

Athlete physical fitness sports training cyber-physical systems internet of things performance optimization

1. Introduction to the physical fitness and cyber-physical system

The role of sports in advancing human civilization has been significant throughout history. India has a long and illustrious history of sports that extends back to ancient times, and to the development of modern sports in India, such as cricket and field hockey [1]. Sports have been an integral aspect of cultural, social, and religious practices in India, encompassing various activities such as wrestling, kabaddi, hockey, cricket, and numerous others. Sports encompass physical exertion and cognitive and affective growth [2]. The sports sector has experienced substantial growth in recent times, owing to the government’s backing and allocation of resources towards enhancing infrastructure, training amenities, and research in sports science. Consequently, there has been a rise in the number of individuals participating in sports, training athletes, and having a keen interest in sports nationwide.

The significance of athletes in sports programs cannot be overstated, as they serve as the foundation for attaining success in the field [3]. The incorporation of physical fitness is a crucial element in enhancing athletic performance. To optimize their performance, minimize the risk of injuries, and expedite the recovery process, athletes must possess exceptional levels of physical fitness. The components of physical fitness encompass muscular strength, cardiovascular endurance, joint flexibility, swiftness, and nimbleness [4]. Enhancing physical fitness necessitates a blend of exercise, dietary, and recuperation approaches.

The optimal performance of athletes is contingent upon their physical fitness and a meticulously crafted training regimen [5]. Physical fitness is a fundamental aspect of success in sports, as it allows athletes to execute the essential skills, techniques, and tactics integral to their respective athletic disciplines. The training program is a methodical regimen designed to enhance athletes’ physical and mental capacities, enabling them to achieve excellence in their sports.

The attainment of physical fitness is associated with numerous advantages, including improving endurance, strength, speed, agility, flexibility, and coordination [6]. Various sports require specific physical attributes, and athletes possessing superior physical fitness enjoy a competitive advantage over their peers. A meticulously crafted training regimen is imperative to guarantee that athletes can sustain or enhance their physical fitness capacity.

The training regimen must be customized to suit the athlete’s particular sport, level of physical fitness, and desired outcomes [7]. The program should encompass exercises that prioritize the development of muscular power, cardiovascular stamina, and range of motion. Athletes must engage in targeted activities replicating the movements necessary for their respective sports to enhance their technique and coordination.

Several obstacles are encountered in sports training when it comes to enhancing the physical fitness of athletes in conventional systems [8, 9, 10].

•
The absence of personalized training regimens is familiar with conventional training programs, which results in overtraining or undertraining among athletes.
•
The insufficiency of monitoring and feedback challenges coaches to modify training programs as required due to the limited information on an athlete’s progress.
•
Insufficient measures for injury prevention: Conventional training programs must sufficiently cater to injury prevention, thereby increasing the likelihood of injuries.
•
Inadequate recovery time can impede athletic performance and increase the likelihood of injury among athletes.
•
The utilization of technology in training programs is restricted, impeding the potential for enhanced performance and training.
•
The limited availability of resources results in discrepancies in training and performance among athletes residing in less affluent regions compared to their counterparts in more privileged areas.

The proposed method should have the potential to overcome the difficulties above by offering unbiased real-time, individualized physical fitness assessments. Using Internet of Things (IoT) sensors and Cyber Physical System (CPS) enables the uninterrupted monitoring of an athlete’s physical activity, which can be scrutinized to furnish valuable insights into their performance and identify potential areas for enhancement. This approach can assist coaches and trainers in developing training programs tailored to individual athletes’ unique needs and goals, enhancing their efficacy.

The primary contributions are listed below:

•
This study proposes a physical fitness evaluation system for athletes called IoT, CPS-based Physical Fitness Evaluation Method (IoT-CPS-PFEM).
•
The proposed method used loT-based prediction model, Cluster based prediction model, and Fuzzy support vector regression to predict athletes’ fitness in training.
•
The present study involves the utilization of MATLAB/Simulink for the analysis of the Physionet Challenge 2017 dataset.

The paper is structured as follows: Section 2 furnishes a comprehensive overview of the background information and literature survey about evaluating athlete physical fitness through the employment of the IoT and CPS. In Section 3, an IoT-CPS-PFRM is suggested to assess the physical fitness of athletes during training. This framework utilizes a fuzzy system and clustering algorithm. In Section 4, a simulation analysis is performed using MATLAB/Simulink on the proposed method, utilizing the Physionet Challenge 2017 dataset. Section 5 presents the conclusions and outlines the potential areas for future research regarding the proposed methodology.
2. Background and literature survey

The present study encompasses a comprehensive literature review about the analysis and development of athlete fitness, focusing on research studies to optimize athletes’ physical fitness and performance. This paper examines diverse methodologies and strategies employed in evaluating athlete fitness and evaluates the efficacy of distinct training regimens in enhancing their physical fitness capacities.

The study conducted by Moseid et al. investigated the correlation between the physical fitness level of elite youth athletes and the frequency and severity of injury and illness [11]. The research encompassed a cohort of 607 athletes from diverse sporting disciplines. It revealed that individuals with elevated levels of physical fitness experienced a reduction in both the frequency and severity of injuries and illnesses. The research emphasizes the significance of maintaining physical fitness as a preventive measure against injuries and diseases among adolescent athletes.

The study conducted by Henriques-Neto et al. aimed to assess the testing-retesting reliability of athlete physique tests administered through the FITescola^® battery among a cohort of young athletes [12]. The research encompassed a total of 56 athletes, and the findings indicated good test-retest reliability for the majority of the assessments. The FITescola^® battery has demonstrated high reliability in evaluating physical fitness among adolescent athletes, rendering it a valuable assessment instrument.

The study by Mathisen et al. aimed to examine the athlete health and conditions of stamina shortage in women athletes [13]. The research encompassed a cohort of 57 female fitness athletes. The findings indicated that most athletes exhibited relative energy deficiency, which could adversely affect their overall well-being and athletic capabilities. The research highlights the significance of vigilantly tracking the balance between energy intake and expenditure among female fitness athletes to safeguard their well-being and optimize their athletic performance.

In their recent publication, Gao et al. introduced an enhanced random forest algorithm and a model based on fuzzy mathematics to evaluate the physical fitness of athletes [14]. The research encompassed a cohort of 150 sportspeople, and the findings indicated that the suggested framework exhibited superior performance in terms of precision and effectiveness compared to alternative frameworks. The model that has been presented displays potential applications in the evaluation and forecasting of physical fitness for athletes.

The study by Chang et al. aimed to compare the physical fitness, Functionality Movement Screening (FMS), and Star Excursing Balanced Test (SEBT) results of junior athletes with varying risks of sports injuries [15]. The research encompassed a cohort of 197 adolescent athletes representing various sports disciplines. The findings indicated that individuals who exhibited elevated scores in both FMS and SEBT demonstrated superior levels of physical fitness. The research emphasizes the significance of evaluating movement quality and balance in the context of injury prevention and physical fitness promotion for young athletes.

Yuan et al. have presented a research paper outlining a neural network system that utilizes motion sensors to assess the physical fitness of basketball players [16]. The system collects data about movement through the utilization of sensors and is subjected to processing via neural network algorithms. The simulation findings indicate that the suggested approach can effectively assess the physical aptitude of basketball athletes. Notwithstanding, the investigation exhibits certain constraints, including the restricted sample size and the necessity for additional verification through an expanded dataset.

Li et al. presented an intelligent sports training system that utilizes artificial intelligence and big data [17]. The utilization of data analytics, machine learning, and cloud computing is integrated into the system to enhance the efficiency of athlete training programs. The system exhibited encouraging outcomes in improving athletes’ physical fitness and performance.

Xu et al. presented a model for classifying sports training videos using deep learning techniques [18]. The system employs Convolutional Neural Networks (CNNs) to categorize training videos according to their content. This facilitates the process of coaches selecting suitable training materials for their athletes. The proposed method has demonstrated a high level of accuracy in the classification of training videos. Nonetheless, the investigation needed to assess the efficacy of the suggested framework on a comprehensive dataset, thereby constraining the applicability of the findings.

A CPS framework was proposed by Arafsha et al. to measure and analyse physical activities [19]. The proposed system incorporates wearable sensors, IoT devices, and cloud computing to gather and evaluate data related to physical activity. The proposed framework demonstrated exceptional precision in identifying physical activities, rendering it a valuable tool in diverse sports training contexts. Nevertheless, the efficacy of the system in practical scenarios necessitates additional assessment.

Bhaumik et al. have presented a paper outlining a suggested Activity Recognition in Video Surveillance (ARVS) system for recognizing incidents and operations within the context of CPS [20]. Using computer vision and machine learning methodologies enables the system to identify and classify various events and actions depicted in video recordings. The method proposed in this study has demonstrated high accuracy in recognizing diverse events and activities. As a result, it holds potential for application in scenarios such as sports training and surveillance. Nonetheless, the manuscript should have addressed any constraints or possible obstacles when deploying the suggested framework in practical systems [21].

The literature review has identified several challenges associated with conventional sports training for enhancing athletes’ physical fitness. These challenges include the absence of precise and objective fitness assessment techniques, inadequate monitoring of physical activities, and insufficient customized training programs [22]. The papers under review presented diverse approaches to tackle the challenges above. These approaches encompassed a motion sensor-driven neural network for evaluating basketball physical fitness. This smart athlete training system relied on machine learning methods, and a video classification model for sports training based on deep learning. Nevertheless, these techniques were restricted by exorbitant expenses, intricacy, and insufficient practicality. Consequently, it is imperative to develop an evaluation method for physical fitness based on IoT and CPS that amalgamates the advantages of CPS, IoT, and data analytics to ensure precise and individualized fitness assessment and training. The IoT-CPS-PFEM has the potential to address the shortcomings of conventional techniques and offer a thorough and effective approach for assessing and enhancing athlete physical fitness in sports training settings.

3. Proposed IoT-CPS-PFEM

The IoT-CPS-PFEM is a new approach for assessing athletes’ physical fitness while training. IoT and CPS technologies collect data from sensors like heart rate monitors, accelerometers, and GPS. Next, machine learning methods including the IoT-based prediction model, cluster-based prediction model, and fuzzy support vector regression are used to analyse the data and estimate athletes’ physical fitness levels. This study aims to develop a reliable and effective method for assessing athletes’ physical aptitude and providing rapid responses to optimize their training regimen and improve their athletic talents. The suggested approach uses IoT and CPS technologies to monitor athletes’ physical activity in real-time and provide personalized training schedules based on their fitness levels. This strategy reduces injury risk and optimizes athlete performance. The system comprises multiple interconnected blocks that collaborate to attain the intended objective. An in-depth examination of each of them will be conducted.

Cloud storage server: The topic of discussion pertains to a server utilized for cloud storage. The function of this particular block relates to the storage of data obtained from diverse sources, including wearable sensors, mobile devices, and other IoT-enabled devices. The cloud storage service offers a scalable, secure, and dependable solution for storing and managing vast quantities of data.

Dataset selection: The block responsible for dataset selection retrieves pertinent data from the cloud storage server and supplies it to the Physical Fitness (PF) – CPS algorithm block. This module minimizes extraneous data and enhances the system’s efficiency.

Physical fitness – CPS algorithm: The PF-CPS algorithm is a computational procedure to assess an individual’s physical fitness. The fundamental component of the system employs machine learning algorithms to scrutinize data and furnish discernments regarding the athletic prowess of individuals in training environments. The PF-CPS algorithm considers multiple physiological variables, like heart rate, blood temperature and pressure, oxygen saturation, and fitness level, to assess an athlete’s level of physical fitness.

Cyber-physical system: The CPS module comprises a range of IoT-enabled gadgets, including wearable sensors, mobile devices, and other health-monitoring equipment. These devices acquire real-time data and transmit it to a cloud-based storage server for analysis.

IoT-based CPS (IoT-CPS) in Cloud: The IoT-CPS in Cloud module executes the IoT-CPS algorithm within a cloud-based setting. The present module has been devised to effectively manage substantial volumes of data and furnish instantaneous observations about the athlete’s physiological well-being in training environments.

The system utilizes IoT and CPS technologies to gather, analyse, and process real-time data from diverse sources, offering valuable insights into athletes’ physical well-being shown in Fig. 1.

Figure 1.

System architecture of the proposed IoT-CPS-PFEM.

3.1 IoT architecture

It is predicated on the combination and cloudification of extant databases. Consequently, every business system must have a consolidated entry and management system. A hierarchical approach comprising a data collection, processing, and comprehensive application layer is recommended for its structure.

3.1.1 Data gathering layer

Utilize the data acquisition methodology of the pre-existing model. Data is collected via diverse equipment for data collection, including automatic detection, wireless communications, Local Area Networks (LAN), Wifi, GPS, and other data collection methods.

3.1.2 Data processing layer

The data processing layer is a component of a system that is responsible for handling and manipulating data. The layer responsible for gathering data on a significant scale is the data collection layer. The data is then consolidated to facilitate comprehensive evaluation and informed decision-making. This enables the system to achieve surveillance, pattern evaluation, early alarm, and other related functions in training environments.

3.1.3 Comprehensive application layer

The coordinated administration of the integrated gateway for the IoT. Various categories and tiers of data processing systems necessitate distinct sign-in or verification techniques. Achieve the management and regulation of all commercial systems using the IoT and the next layer. Securing sign-in involves diverse techniques like passwords, 2FA, MFA, and OAuth. Identity Federation integrates external providers, while features such as device recognition and behavioral biometrics enhance security. Combining these ensures a robust sign-in approach across system categories.

The IoT-CPS-PFRM approach proposes an optimization strategy involving the dynamic allocation of cloud resources, based on IoT applications’ quality and latency requirements. This is achieved by applying a Fuzzy Support Vector Regression model, which considers various factors, including energy consumption, response time, and resource utilization. These metrics play a pivotal role in providing a nuanced understanding of the model’s operational efficiency and effectiveness in practical applications.

Figure 2.

Functional distribution of the IoT-CPS-PFEM.

Figure 2 illustrates that the functional dispersion is categorized into three layers: application layer, end user layer, and central cloud service. The system has been designed to incorporate a cohesive transmission and information structure. Unifying sub-systems, module designs, and methods within each layer are advised to minimize the number of classes and overall complications.

Central cloud service: It comprises two primary systems: applications and data services. The primary function is to store and oversee all gathered data, promptly evaluate and handle the existing data, and provide outcomes, evaluations, and recommendations in training environments.

The data servicing method processes the gathered large-scale data in compliance with the prescribed evaluation regulations. Establishing a dynamic background data cleaning docker cluster is imperative, as it enables the scaling of docker by the fluctuating volume of data requiring cleansing. Effectively managing large-scale data within a central cloud service involves a structured approach. The process begins with data ingestion, where information from diverse sources is collected and securely transported to the cloud.

Surrounding fog service nodes: The function is to gather data obtained by detectors and peripheral fog terminals, store it promptly, and transmit it expeditiously to the main cloud service.

Surrounding fog terminal: Mobile applications are designed to cater to users’ needs by collecting data that requires manual input, analysing this data along with data received, and providing practical outcomes and relevant alarms. Furthermore, users can access the applications offered by the main cloud to procure personalized and all-encompassing reports and diagnostic recommendations.

3.2 Physical fitness forecasting

The proposed IoT-CPS-PFRM method employs clustering to categorize comparable IoT devices and applications according to their attributes and needs. This approach facilitates the effective allocation and administration of resources. The objective is to enhance the system’s efficiency by minimizing energy usage and enhancing response time via proficient clustering.

Forecasting athletes’ physical fitness can be categorized as a form of data gathering and analysis. The clustering algorithm partitions the items within a dataset into various clusters or categories to extract valuable insights [23]. The process of data clustering involves grouping data objects in a manner that maximizes their similarity within a cluster while minimizing the similarity between different sets. The utilization of data clustering holds significant value in various domains, including but not limited to the recognition of patterns, retrieval of information, e-commerce, advertising, and document categorization using grid clustering. The grid clustering method involves dividing the data space into a finite number of cells or grids of uniform sizes. The data points are clustered into the closest grid cells based on their proximity to the grid center. Considering the data within a group as a unified entity in various applications is possible. Clustering can facilitate the identification of dense or sparse regions, thereby enabling the discovery of global dispersion trends and significant relations among data elements. The area of data analysis has been the subject of research aimed at identifying suitable techniques for efficient and feasible cluster analysis of extensive records. The article centers on the efficacy of intricate clustering shapes, diverse data types, and clustering approaches designed for mixed categorical and numerical information in extensive datasets in training environments.

The conventional approach to clustering involves initially establishing the distance metric between entities, followed by applying a suitable algorithm to group the entities based on the computed distance metric. This topic’s categorization includes division, hierarchical, density, and grid-based techniques. The classification can be broadly classified into five distinct methods: division, hierarchy, density, grid, and model-based. The prevalent methods for measuring distance are Euclidean distance, Manchester distance, and Mincos distance. Equations (1) to (3) are presented in the Euclidean distance, Manchester distance, and Mincos distance.

$\displaystyle d\left({x,y}\right)=\sqrt{\left|{i_{x_{1}}-i_{y_{1}}}\right|^{2}% +\left|{i_{x_{2}}-i_{y_{2}}}\right|^{2}+\vartheta+\left|{i_{x_{p}}-i_{y_{p}}}% \right|^{2}}$ (1) $\displaystyle d\left({x,y}\right)=\sqrt{\left|{i_{x_{1}}-i_{y_{1}}}\right|+% \left|{i_{x_{2}}-i_{y_{2}}}\right|+\vartheta+\left|{i_{x_{p}}-i_{y_{p}}}\right|}$ (2) $\displaystyle d\left({x,y}\right)=\sqrt{\left|{i_{x_{1}}-i_{y_{1}}}\right|^{q}% +\left|{i_{x_{2}}-i_{y_{2}}}\right|^{q}+\vartheta+\left|{i_{x_{p}}-i_{y_{p}}}% \right|^{q}}$ (3)

The two-dimensional data input is denoted $i_{x_{k}}$ and $i_{y_{k}}$ . The calculation relationship for the weighted distance can be expressed in Eq. (4), where each variable is assigned a weight $w_{x}$ based on its significance.

$\displaystyle d\left({x,y}\right)=\sqrt{w_{1}\mbox{?}\left|{i_{x_{1}}-i_{y_{1}% }}\right|^{2}+w_{2}\left|{i_{x_{2}}-i_{y_{2}}}\right|^{2}+\vartheta+w_{p}\left% |{i_{x_{p}}-i_{y_{p}}}\right|^{2}}$ (4)

The two-dimensional data input is denoted $i_{x_{k}}$ and $i_{y_{k}}$ . The assigned weight variable is denoted $w_{x}$ .

3.3 Fuzzy Support Vector Regression (FSVR)

FSVR is a machine learning technique that integrates support vector regression and fuzzy logic to effectively manage uncertainties in the data. FSVR combines support vector regression and fuzzy logic for effective uncertainty management in data. Strategies include integrating fuzzy logic membership functions, fuzzy partitioning, and adaptive fuzzy inference systems. The IoT-CPS-PFRM employs this tool to forecast the performance of IoT applications in diverse scenarios and to enhance resource allocation strategies based on the anticipated performance.

The provided data consists of a training sample $S=\left\{{\left({i_{1},j_{1}}\right),\left({i_{2},j_{2}}\right),\ldots,\left({% i_{n},j_{n}}\right)}\right\}$ , which is comprised of n elements in the form of ( $i_{x}$ , $j_{x}$ ). Here, $i_{x}$ represents the feature vector of the x-th sample, while $j_{x}$ denotes the label associated with the model. The expression for the categorization hyperplane in the sample space is shown in Eq. (5).

$\displaystyle W^{T}i+b=0$ (5)

The formula involves the utilization of an average vector, denoted as $W$ , which establishes the direction of the hyperplane. The formula includes a displacement term, marked as $b$ , which computes the separation between the hyperplanes and the source $\left(i\right)$ . The process of deciding a categorized hyperplane involves the determination of the values of $W$ and $b$ . To achieve accurate sample categorization and establish the categorization interval, it is imperative to ensure that the following relationship is duly satisfied using Eq. (6).

$\displaystyle j_{x}\left\{{W^{T}i_{x}+b}\right\}\geqslant 1;x=1,2,3,\ldots,n$ (6)

The weight and biasing condition are denoted $W$ and $b$ . The initial fitness data is expressed $i_{x}$ , and the relationship is expressed $j_{x}$ . The objective of identifying the optimal categorization hyperplane is to optimize the total distance. Consequently, determining the optimal hyperplane can be reformulated using Eqs (7) and (8).

$\displaystyle\min\left\{{\frac{1}{2}\left|W\right|^{2}}\right\}>0$ (7) $\displaystyle j_{x}\left\{{W^{T}i_{x}+b}\right\}\geqslant 1$ (8)

The concept of soft margin is employed to mitigate the impact of anomalous points on the categorization hyperplane and prevent overfitting of the model. The weight and biasing condition are denoted $W$ and $b$ . The initial fitness data is expressed $i_{x}$ , and the relationship is expressed $j_{x}$ . This involves incorporating slack elements regulated by the constant $K$ to manage the extent of errors. The connection of $\min\left\{{\frac{1}{2}\left|W\right|^{2}}\right\}$ is expressed in Eq. (9).

$\displaystyle\min\left\{{\frac{1}{2}\left|W\right|^{2}}\right\}+K\sum_{x=0}^{n% -1}\propto_{x};\propto>0$ (9)

The conventional support vector machine needs to improve its accuracy of models in the presence of noise points due to the equal weighting assigned to each sample point in the categorization hyperplane. The weight is $W$ , the Lagrange factor is $\propto_{x}$ , and the error constant is $K$ . The Fuzzy Support Vector Machine (FSVM) is a computational model that integrates the concept of membership of fuzzy with a support vector machine. Using a membership operation, this approach assigns a membership value, denoted as $M_{x}$ , to each sample point in training environments. The variation in membership degree among sample points results in distinct slack variables, thereby mitigating the issue of noise point interference and enhancing the model’s predictive capacity. The Lagrange membership function has been established using Eq. (10).

$\displaystyle f\left(i\right)=\sum_{x=0}^{n-1}\left({\propto^{\prime}_{x}}% \right)k\left({i,i_{x}}\right)+b$ (10)

The Lagrange coefficient is $\propto^{\prime}_{x}$ , and the error constant is $k\left({i,i_{x}}\right)$ , where the input is $i_{x}$ , and the reference input is $i$ . The biasing weight is $b$ . A physical fitness measure of time can be utilized to express the membership level, as shown in Eq. (11).

$\displaystyle M_{x}=f\left({t_{x}}\right)$ (11)

The computation time is denoted $t_{x}$ , and the fitness function is $f$ . The present study aims to prioritize using data near the prediction point to enhance the accuracy of regression forecasting in the research procedure. The quadratic formula can express membership functioning, as shown in Eq. (12).

$\displaystyle M_{x}=\left({1-\delta}\right)\left({\frac{t_{x}-t_{0}}{t_{0}-t_{% 1}}}\right)+\delta$ (12)

The error in the output is denoted $\delta$ , the initial time is $t_{0}$ , the setup time is $t_{1}$ , and the present time at which the fitness is computed is $t_{x}$ . The assessment of predictive outcomes necessitates the consideration of two primary factors: firstly, the degree to which the observed value aligns with the forecasting period, and secondly, the minimization of the forecasting interval’s scope. Two evaluation measurements, coverage, and width can be stated based on the two conditions. Equations (13) to (15) articulate the connection between them.

$\displaystyle\textit{Coverage}=\frac{1}{n}\sum_{x=0}^{m-1}c_{x}$ (13) $\displaystyle c_{x}=\left\{c_{x}=\begin{cases}1&\bar{i}_{x}^{\textit{low }}<i_% {x}<\bar{i}_{x}^{\textit{up }}\\ 0&\textit{else }\end{cases}\right.$ (14) $\displaystyle\textit{width}=\frac{1}{m}\sum_{x=0}^{m-1}\sqrt{\left(\bar{i}_{x}% ^{\textit{low }}-i_{x}\right)^{2}+\left(\bar{i}_{x}^{up}-i_{x}\right)^{2}}$ (15)

The lower and upper bound of the fitness goal are denoted $\bar{i}_{x}^{\textit{low }}$ and $\bar{i}_{x}^{\textit{up}}$ . The coverage is found using the fitness covering function $c_{x}$ , and the width is used to find the level at which membership function matches with fitness function. The present study employs the techniques above to investigate the system and index maximisation of forecasting methods in training environments for athletes within the context of the IoT [25]. The study utilizes the IoT, physical fitness forecasting, prediction modelling, index optimization, and related methodologies.

Figure 3.

Fitness prediction model of the IoT-CPS-PFEM.

The fitness prediction model of the IoT-CPS-PFEM is shown in Fig. 3. The methods have models IoT-based, cluster-based, and fuzzy support vector regression models. The physical fitness values are computed using subblocks in each method.

3.4 Evaluation system

3.4.1 Factors of physical fitness

Physical fitness is influenced by various factors, including but not limited to muscular strength, stamina, and balance. Engaging in physical activity is a beneficial practice that promotes maintaining physical and cognitive health. Regular physical activity, including cardiovascular exercises and strength training, yields substantial benefits for both physical and cognitive health. It enhances cardiovascular well-being, muscle strength, and weight management, while also improving mental health, reducing stress, and promoting better sleep. The present study examines the physical fitness of individuals in training environments who engage in sports. In assessing physical fitness, it is imperative to analyse its fundamental components, including bodily function, configuration, and physical aptitude. The component system of sports incidents is distinguished from other events by its distinctive features, which include sensitivity, resilience, and other physical qualities integral to physical abilities rather than mere physical elements.

3.4.2 Athlete fitness evaluation method

Before implementing the analytic hierarchy process to develop an assessment system for athletes, it is imperative to establish precise goals and identify the specific entities that will be evaluated. This paper focuses on athletes who have achieved outstanding performance in athletic competitions as the subject of research in training environments. It is necessary to explicate the indexing structure and indexing coefficient, finalize the construction of the assessment condition, and institute the athlete assessment mechanism.

The primary objective of this paper is to identify the indicators pertinent to developing a physical fitness assessment system for athletes. To achieve this goal, a comprehensive review of relevant literature and fieldwork was conducted, identifying numerous indicators closely associated with the evaluation of athletes. The present study aims to explicate the key parameters that determine the physical condition of athletes. In this regard, the primary factors of utmost significance are physique, health, and sports quality. Twelve supplementary indicators contribute to the overall assessment of the athlete’s fitness in training environments. The objective is to devise a physical fitness assessment framework for athletes utilizing the Analytic Hierarchy Process (AHP) [25].

3.5 Weight computation of the fitness evaluation

3.5.1 Computation of the indexing coefficient

The square root method is commonly employed for weight computation in the analytic hierarchy procedure. The proposed approach involves computing each line’s product and raising the indexing coefficient to the power of $n$ . The steps delineate the comprehensive process of calculation:

Step 1: The expression for obtaining the development of each row component $M$ on the $B$ decision array is shown in Eq. (16).

$\displaystyle M_{i}=\sum_{j=0}^{n-1}B_{i,j};i=0,1,\ldots,n-1$ (16)

Step 2: Determine the nth power root, denoted as $w$ , of the product $M$ of every row as presented in Eq. (17):

$\displaystyle w_{i}=\sqrt{M_{i}};\text{for}\ i=0,1,\ldots,n-1$ (17)

The variable $n$ denotes the order of the evaluation matrix.

Step 3: involves the computation of the index weight $w$ using Eq. (18).

$\displaystyle W_{i}=\frac{w_{i}}{\sum_{i=0}^{n-1}w_{i}}$ (18)

The weighted index is denoted $w_{k}$ .

3.5.2 Consistency test

Inconsistencies in the assessment matrix arise due to variations in the evaluating abilities and value directions researchers adopt when assessing the metrics. A consistency test’s assessment matrix may contain discrepancies for a number of reasons. Subjectivity is introduced by assessors’ differing interpretations of the evaluation criteria, and inequalities in judgments are a result of differing levels of competence. The issue is more proportionality in assigning grades to critical indications. Hence, it is unfeasible to formulate a coherent matrix when the $N$ opinion matrix’s arrangement surpasses two due to the escalating complexity in reconciling multiple perspectives. Nonetheless, there is a particular discrepancy between the matrix of judgments and the uniformity condition. The consistency test aims to determine the decision matrix’s general acceptance precisely. The procedural stages of the consistency inspection procedure:

Step 1: the value $\propto_{\max}$ (maximum eigenvalue) was obtained from the decision matrix, as indicated by Eq. (19):

$\displaystyle\textit{AW}=\left(\begin{array}[]{cccc}1&{k_{12}}&\ldots&{k_{1m}}% \\ {k_{21}}&1&\ldots&{k_{2m}}\\ \vdots&\vdots&\vdots&\vdots\\ {k_{m1}}&{k_{m2}}&\ldots&1\\ \end{array}\right)\left(\begin{array}[]{c}{W_{1}}\\ {W_{2}}\\ \vdots\\ {W_{m}}\\ \end{array}\right)$ (19)

The matric element is denoted $k_{xy}$ , and the weight value is denoted $W_{i}$ .

Step 2: The consistency indexing, denoted as CI, is computed using Eq. (20).

$\displaystyle\textit{CI}=\frac{\propto_{\max}-m}{m-1}$ (20)

The maximum eigenvalue is $\propto_{\max}$ , and the sample count is $m$ .

Step 3: Compute the Consistency Ratio (CR) metric utilizing Eq. (21).

$\displaystyle\textit{CR}=\frac{\textit{CI}}{\textit{RI}}$ (21)

The variable denoted as RI represents the randomness index. The degree of the consistency index is determined by CI, the quantity of $N$ comparison indices, based on the created at random relevant assessment matrix.

3.6 Athlete fitness evaluation flow

As per the comprehensive layout of the evaluation test administration system, the method will authenticate and validate the users’ credentials attempting to access the framework for investigation and management purposes. Before conducting candidate exams or managing teachers, it is imperative to perform identity verification procedures. Figure 4 illustrates the primary operational sequence of the system. Established procedures for identity verification are pivotal in maintaining exam integrity and managing teachers in the evaluation system. A detailed framework ensures thorough verification steps for individuals participating in exams. On the other hand, managing teachers involves a comprehensive identity verification system, contributing to the overall security and credibility of both exam sessions and the teacher.

Figure 4.

The operation sequence of the proposed IoT-CPS-PFEM.

Upon successful login by the student user, they will be directed to the test reservations page, where the trainer will administer the appropriate examination at the designated reservation time. Upon completion of the study, the instructor will input the corresponding ratings into the framework, generating a test report based on the recorded scores. By individual students’ varying physical health conditions, teachers will provide appropriate fitness direction. Upon entering the accounting software, the teacher could effectively sign in. A thorough health assessments that consider factors like medical history and injuries, educators create personalized fitness plans. These plans include modifications or alternative exercises as needed.

As per the current requirements of educators, it is imperative to incorporate administrative capabilities such as user addition, modification, deletion, and grade administration. Upon completion of the health assessment and diagnosis, it is incumbent upon educators to thoroughly analyse the results and provide appropriate guidance in training environments. Educators analyze health assessments, collaborating with healthcare professionals for insights. Personalized fitness plans are crafted, with modifications for safety. Open communication addresses preferences, and continuous monitoring ensures dynamic fitness guidance in training settings.

Administrative users can execute fundamental functions, such as the addition, removal, and alteration of user accounts for both educators and pupils. Furthermore, users can finalize the configuration and administration of information about the system.

The IoT-CPS-PFEM is a proposed system for evaluating the physical fitness of athletes. It leverages IoT and CPS technologies to forecast the fitness levels of athletes during their training sessions. The integration of diverse prediction models, such as loT-based and cluster-based models, alongside Fuzzy Support Vector Regression, enables the system to provide precise and effective forecasts of athletes’ physical fitness. The system aims to enhance athlete training by furnishing instantaneous feedback on their physical fitness levels and facilitating customized training regimens tailored to their specific requirements. The system employs a grid clustering algorithm to categorize sportspeople based on comparable physiological traits.

4. Simulation analysis and findings

MATLAB/Simulink is extensively utilized for modelling and simulating CPS. The software offers extensive resources for modelling, simulating, and analysing CPS [26]. The IoT – CPS – Physical Field Experimental Method (PFEM) system can be affected. The utilization of MATLAB/Simulink necessitates a minimum of 4 GB of RAM and a dual-core CPU and is compatible with Windows 7 or a more recent version.

The dataset utilized in the Physionet Challenge 2017 is openly accessible and encompasses information obtained from devices with sensors affixed to athletes during diverse physical activities for 40,336 patients [27]. The data set comprises physiological metrics such as heart rate, blood pressure, oxygen saturation level, Global Positioning System (GPS), and accelerometer readings. The corpus comprises over ten thousand audio recordings and an estimated fifty-five gigabytes. Apart from electrocardiogram (ECG) data, certain patients were also subjected to other forms of physiological data collection, including blood pressure, oxygen saturation, and respiratory rate, utilizing diverse sensor types.

The present study proposes utilizing five simulation metrics to assess the efficacy of the IoT-CPS-PFEM approach in MATLAB.

Accuracy – The metric of accuracy evaluates the overall precision of the IoT-CPS-PFEM algorithm when assessing the physical fitness of an athlete. Precision evaluation involves comparing the anticipated values generated by the IoT-CPS-PFEM algorithm and the actual values.

Precision – The precision metric is utilized to assess the IoT-CPS-PFEM algorithm’s accuracy in identifying athletes’ physical fitness level. The precision calculation involves dividing the total number of true positives by the total of true positives and false positives.

Recall – The recall metric is utilized to evaluate the ability of the IoT-CPS-PFEM algorithm to identify an athlete’s physical fitness level. The recall calculation involves dividing the count of true positives by the total of true positives and false negatives. The recall metrics assesses the algorithm’s proficiency in correctly identifying athletes with high physical fitness, ensuring minimal instances of false negatives. Additionally, the recall metric emphasizes the algorithm’s sensitivity, aiming to capture all occurrences of high physical fitness comprehensively.

F1 Score – The F1 Score is a performance metric that provides a balanced evaluation of the IoT-CPS-PFEM algorithm’s precision and recall. It is calculated as a weighted average of these two measures. The fundamental concept of measurement involves the use of ratios as units. The computation of the F1 score consists of utilizing the harmonic average of the precision result and recall result.

Execution Time – The metric of Execution Time pertains to the duration required by the IoT-CPS-PFEM algorithm to assess the physical fitness of an athlete. The computation of the execution time is achieved by measuring the duration between the initiation and termination of the IoT-CPS-PFEM algorithm.

Figure 5.

Accuracy analysis of IoT-CPS-PFEM.

Figure 6.

Precision analysis of the IoT-CPS-PFEM.

The accuracy achieved by various techniques for assessing 1000 athletes’ physical fitness during the training and testing phases are plotted in Fig. 5. The IoT-CPS-PFEM approach demonstrated superior accuracy to current methodologies, achieving 92.2% during training and 90.7% during testing. The enhanced precision observed in the proposed method can be ascribed to the utilization of IoT and CPS, which enables the instantaneous monitoring and analysis of various fitness parameters of athletes, including heart rate, body temperature, and blood pressure. The proposed method’s outcomes have the potential to enhance athlete performance, prevent injuries, and facilitate customized training programs.

Figure 6 displays the precision of diverse techniques employed for assessing the physical fitness of athletes throughout their training and testing phases. The IoT-CPS-PFEM approach demonstrated notable enhancements in accuracy compared to current methodologies, achieving a training accuracy of 94.3% and a testing accuracy of 89.7%. The proposed method integrates the IoT and CPS to facilitate the real-time monitoring of an athlete’s fitness parameters. This integration could enhance the accuracy of the evaluation process. The envisaged approach can yield multiple favorable consequences, such as improved athletic performance, injury mitigation, and the creation of tailored training regimens.

Figure 7.

Recall analysis of the IoT-CPS-PFEM.

The recall of various techniques for assessing athlete fitness during the training and testing phases are shown in Fig. 7. The IoT-CPS-PFEM approach significantly enhanced recall performance compared to current methodologies, achieving a training recall of 92.9% and a testing recall of 90.3%. The proposed method incorporates the IoT and CPS to monitor athletes’ fitness parameters in real-time continuously. This approach has the potential to enhance the accuracy of recall during the evaluation process. The method under consideration has the potential to yield multiple favorable results, such as improved athletic performance, injury mitigation, and the creation of customized training regimens.

Figure 8.

F1 score analysis of the IoT-CPS-PFEM.

Figure 8 indicates the F1 score (%) of different techniques for assessing athlete fitness during the training and testing phases. The IoT-CPS-PFEM approach exhibited an enhancement in F1 score in contrast to pre-existing methodologies, achieving a score of 89.5% during the training phase and 87.5% during the testing phase. The proposed method integrates the IoT and CPS to facilitate the real-time monitoring of an athlete’s fitness parameters. This approach has the potential to enhance the F1 score during evaluation. The suggested method can yield multiple favorable results, such as improved athletic performance, mitigation of injuries, and the creation of tailored training regimens.

Figure 9.

Execution time analysis of the proposed IoT-CPS-PFEM.

Figure 9 summarizes the IoT-CPS-PFEM approach’s efficacy for evaluating athlete fitness using IoT and CPS. This approach is compared to ten other reference techniques for accuracy, precision, recall, F1 score, and execution time for the training and testing phases. The findings indicate that the method exhibited superior performance compared to all the reference methods regarding the accuracy, precision, recall, and F1 score for both the training and testing phases. Moreover, the suggested methodology exhibited a noteworthy enhancement in the duration of execution for both the training and testing phases in contrast to the benchmark techniques. The findings of the proposed process suggest that integrating IoT and CPS methods can significantly enhance the efficacy of athlete fitness assessment techniques, resulting in more precise and streamlined outcomes.

The study presents simulation results that compare the performance of the suggested IoT-CPS-PFRM with ten other methods for an athlete fitness assessment. The technique that has been offered exhibits noteworthy enhancement in all metrics for both the training and testing phases. The proposed methodology attains a training accuracy of 92.2%, precision of 94.3%, recall of 92.9%, and F1 score of 89.5%. Similarly, for testing, the methodology achieves an accuracy of 90.7%, precision of 89.7%, recall of 90.3%, and F1 score of 87.5%. The method under consideration exhibits superior performance compared to alternative approaches, as evidenced by its execution time of 114.7 seconds for training and 18.4 seconds for testing. In summary, the findings indicate that the IoT-CPS-PFRM presents a viable strategy for assessing athlete fitness by utilizing IoT and CPS technologies. This approach offers superior precision and expedited processing in comparison to current methodologies.

5. Conclusion and future study

The significance of physical fitness for athletes lies in its impact on their performance and susceptibility to injuries. The utilization of CPS and IoT technologies has witnessed a surge in monitoring the physical activities of athletes and offering immediate feedback, thereby facilitating the optimization of their training regimen and enhancing their performance. The present study introduces an approach that combines the IoT, CPS, and Physical Fitness and Rehabilitation Monitoring (IoT-CPS-PFRM) to assess the physical fitness of athletes during training. The proposed methodology entails the acquisition of data from various sensors, including heart rate monitors, accelerometers, and GPS, followed by applying machine learning algorithms to scrutinize the data and furnish discernments regarding the athletes’ physical fitness levels.

The simulation outcomes indicate that the method exhibits higher performance compared to the traditional methods regarding precision, recall, F1 score, and execution time for both training and testing. The technique attained a precision rate of 94.3% during training and 89.7% during testing. It achieved a recall rate of 92.9% during the training stage and 90.3% during the testing stage. The F1 score for the training phase was 89.5%, while for the testing phase, it was 87.5%. The reduction in execution time, compared to the pre-existing methods, assumes significance in real-time applications.

Several obstacles must be surmounted, including data privacy, interoperability, and standardization, to fully accept CPS and IoT technologies in assessing athlete fitness. Future studies concentrate on constructing an all-encompassing framework that amalgamates various sensors and analytics instruments for assessing physical fitness in athletes, guaranteeing compatibility, and tackling apprehensions regarding data confidentiality.

Funding

The authors report no funding.

Data availability statement

No datasets were generated or analyzed during the current study.

Author contributions

All authors contributed to the design and methodology of this study, the assessment of the outcomes and the writing of the manuscript.

Footnotes

Conflict of interest

The authors have no conflicts of interest to report.

References

Parry

. E-sports are not sports. Sport, Ethics and Philosophy. 2019 Jan 2; 13(1): 3-18.

Baceviciene

Jankauskiene

Emeljanovas

. Self-perception of physical activity and fitness is related to lower psychosomatic health symptoms in adolescents with unhealthy lifestyles. BMC Public Health. 2019 Dec; 19: 1-1.

Caldemeyer

Brown

Mulcahey

. Neuromuscular training for the prevention of ankle sprains in female athletes: a systematic review. The Physician and Sportsmedicine. 2020 Oct 1; 48(4): 363-9.

Anzelmo,

. Principles of Exercise and Rehabilitation. In The Sports Rehabilitation Therapists’ Guidebook, 2021, pp. 137-156. Routledge.

Filho

Dobersek

Husselman

. The role of neural efficiency, transient hypofrontality and neural proficiency in optimal performance in self-paced sports: a meta-analytic review. Experimental Brain Research. 2021 May; 239(5): 1381-93.

Witkower

Mercadante

Tracy

. How affect shapes status: Distinct emotional experiences and expressions facilitate social hierarchy navigation. Current Opinion in Psychology. 2020 Jun 1; 33: 18-22.

Cadenas-Sanchez

Migueles

Esteban-Cornejo

Mora-Gonzalez

Henriksson

Rodriguez-Ayllon

Molina-García

Löf

Labayen

Hillman

Catena

. Fitness, physical activity and academic achievement in overweight/obese children. Journal of Sports Sciences. 2020 Apr 2; 38(7): 731-40.

Johnson

Kroshus

Hafferty

Senturia

Garrett

Tandon

. Improving use of physical fitness testing data in middle schools to inform equitable school-wide physical activity practices: a mixed-methods approach. American Journal of Health Education. 2023 Jan 2; 54(1): 50-61.

Joensuu

Kujala

Kankaanpää

Syväoja

Kulmala

Hakonen

Oksanen

Kallio

Tammelin

. Physical fitness development in relation to changes in body composition and physical activity in adolescence. Scandinavian Journal of Medicine & Science in Sports. 2021 Feb; 31(2): 456-64.

10.

Barbosa

Saint Martin

Soares

Fontana

Lan

Kales

Molina

Porto

. The effects of a 6-month mandatory military police academy training on recruits’ physical fitness. Work. 2022 Jan 1; 73(4): 1297-306.

11.

Moseid

Myklebust

Slaastuen

Bar-Yaacov

Kristiansen

Fagerland

Bahr

. The association between physical fitness level and number and severity of injury and illness in youth elite athletes. Scandinavian Journal of Medicine & Science in Sports. 2019 Nov; 29(11): 1736-48.

12.

Henriques-Neto

Minderico

Peralta

Marques

Sardinha

. Test–retest reliability of physical fitness tests among young athletes: The FITescola battery. Clinical Physiology and Functional Imaging. 2020 May; 40(3): 173-82.

13.

Mathisen

Heia

Raustøl

Sandeggen

Fjellestad

Sundgot-Borgen

. Physical health and symptoms of relative energy deficiency in female fitness athletes. Scandinavian Journal of Medicine & Science in Sports. 2020 Jan; 30(1): 135-47.

14.

Gao

Liu

. Application of improved random forest algorithm and fuzzy mathematics in physical fitness of athletes. Journal of Intelligent & Fuzzy Systems. 2021 Jan 1; 40(2): 2041-53.

15.

Chang

Chou

Chang

Chen

. Comparison of functional movement screen, star excursion balance test, and physical fitness in junior athletes with different sports injury risk. BioMed Research International. 2020 Mar 25; 2020.

16.

Yuan

Kamruzzaman

Shan

. Application of motion sensor based on neural network in basketball technology and physical fitness evaluation system. Wireless Communications and Mobile Computing. 2021 Apr 14; 2021: 1-11.

17.

Cui

. Intelligent sports training system based on artificial intelligence and big data. Mobile Information Systems. 2021 May 22; 2021: 1-11.

18.

. A sports training video classification model based on deep learning. Scientific Programming. 2021 May 29; 2021: 1-11.

19.

Arafsha

Laamarti

El Saddik

. Cyber-physical system framework for measurement and analysis of physical activities. Electronics. 2019 Feb 22; 8(2): 248.

20.

Bhaumik

Jana

Mohanta

. Event and activity recognition in video surveillance for cyber-physical systems. InEmergence of Cyber Physical System and IoT in Smart Automation and Robotics: Computer Engineering in Automation 2021 May 5 (pp. 51-68). Cham: Springer International Publishing.

21.

Huang

Muthu

Gayathri

. Wearable energy efficient fitness tracker for sports person health monitoring application. Journal of Intelligent & Fuzzy Systems. 2021(Preprint): 1-10.

22.

Qiu

Liu

Muthu

Sivaparthipan

. Design of an energy-efficient IoT device with optimized data management in sports person health monitoring application. Transactions on Emerging Telecommunications Technologies. 2022 Oct; 33(10): e4258.

23.

Yang

. A deep learning and clustering extraction mechanism for recognizing the actions of athletes in sports. Computational Intelligence and Neuroscience. 2022 Mar 24; 2022.

24.

Zhang

Fei

Shakeel

Samuel

. WITHDRAWN: Computational efficient wearable sensor network health monitoring system for sports athletics using IoT. Aggression and Violent Behavior. 2020

25.

Liu

Chen

Huang

. Constructing an artificial intelligence strategy algorithm for the identification of talented rowing athletes. Soft Computing. 2023 Feb; 27(3): 1743-50.

26.

https://in.mathworks.com/products/simulink.html.

27.

https://peterhcharlton.github.io/info/datasets/cinc2017.