A Wearable System for Real-Time Posture Monitoring and Feedback during Strength Training

MelisaEfuaAnderson1

JoshuaJerrySelormYegbe1

MichaelGyan1

Dr.

HephziTagoe1,2✉Emailhephzi.tagoe@acity.edu.gh

1Academic City UniversityAccraGhana

2Department of Biomedical EngineeringAcademic City UniversityAccraGhana

Melisa Efua Anderson1, Joshua Jerry Selorm Yegbe1, Michael Gyan1, Hephzi Tagoe*1.

1. Academic City University, Accra, Ghana.

*Correspondence: Dr. Hephzi Tagoe, Department of Biomedical Engineering, Academic City University, Accra, Ghana. Email: hephzi.tagoe@acity.edu.gh

Abstract

Lower back pain (LBP) affects an estimated 75–80% of individuals worldwide, with poor posture during strength training identified as a significant contributing factor. This study presents an intelligent, low-cost wearable system for real-time lumbar posture monitoring, demonstrated using Romanian Deadlifts as a case study. The system combines an MPU-9250 inertial measurement unit (IMU), an ESP32 microcontroller, a cloud-deployed Random Forest model (PostureProML), and a Flutter-based mobile application (PostureProne). It achieved a 94.5% classification accuracy across three posture categories; proper, rounded, and arched with minimal angular drift (1.8°–7.1°) and a 4-hour operational battery life. Usability testing with ten participants (aged 18–25) indicated high acceptance, with 90% finding the app intuitive. By enabling immediate feedback and encouraging posture correction, this interdisciplinary solution offers a practical pathway to reducing gym-related injuries.

Keywords:

Wearable technology

inertial measurement unit (IMU)

machine learning

posture monitoring

strength training

Introduction

Musculoskeletal disorders, including posture-related conditions, can develop at any stage of life—during childhood, adulthood, aging, occupational activities, or athletic participation[1–3]. Research has shown that early intervention significantly enhances the healing process, with recovery being notably more effective during childhood due to greater physiological adaptability. With advancements in technology, spinal abnormalities can now be addressed through wearable posture monitoring systems, offering a non-invasive and continuous method for early detection and correction[4–6]. However, as noted by Pandian and Mohanavelu [7, 8], traditional sensors and medical devices often present challenges for long-term physiological monitoring due to discomfort and lack of wearability, highlighting the need for more user-friendly wearable alternatives.

Musculoskeletal disorders, particularly lower back pain (LBP), remain a leading cause of disability and reduced physical performance worldwide, affecting approximately 75–80% of individuals at some point in their lives, with 10–15% developing chronic conditions [9, 10]. Strength training exercises, particularly deadlifts and squats, exert significant compressive and shear forces on the lumbar spine with forces that can lead to injury if not performed with proper biomechanical form or adequate supervision [11, 12]. Correct lumbar alignment is critical to mitigating these risks, yet amateur and unsupervised gym users often struggle to maintain safe posture due to the lack of immediate biomechanical feedback. Existing posture monitoring tools, including wall mirrors, fixed gym cameras, and form-check mobile applications, offer limited precision and delayed feedback, making them inadequate for real-time correction [13, 14].

Most wearable systems rely on inertial measurement units (IMUs) as their core sensing technology [15–17]. These devices enable both discrete and continuous posture monitoring in real-world environments [18, 19], allowing for the rapid detection of static and dynamic postural deviations

that may otherwise go unnoticed when sustained over long periods [20–22]. In addition to their sensing capabilities, wearables offer key advantages that make them well-suited for work activity monitoring: they are cost-effective, lightweight, compact, portable, and energy-efficient [23]. Moreover, they provide objective, reliable, and accurate data [24], supporting realistic and trustworthy assessments of occupational and ergonomic conditions [25, 26]. Recent advances in wearable sensing technologies and embedded machine learning algorithms offer promising avenues for personalised, real-time posture monitoring. Inertial measurement units (IMUs), when integrated with low-power microcontrollers, can accurately track three-dimensional movement patterns during dynamic exercise routines [27]. When coupled with on-device or cloud-based classification models, these systems have demonstrated the potential to support injury prevention in rehabilitation and sports settings [28, 29].

This paper presents an interdisciplinary prototype developed through the integration of hardware and software for intelligent lumbar posture monitoring. The system comprises a wearable IMU-based device featuring an MPU-9250 sensor and ESP32 microcontroller, a Random Forest-based machine learning model (PostureProML), and a Flutter-based mobile application (PostureProne). Together, these components deliver real-time posture classification and personalised feedback to users performing RDLs. The prototype addresses key limitations in existing gym-specific monitoring tools by providing an accessible, low-cost, and scalable solution.

Materials and Methods

Overview of System Architecture

The system (Fig. 1) comprises three subsystems: (1) a wearable unit with an MPU-9250 IMU and ESP32-WROOM-32 microcontroller, (2) a cloud-hosted Random Forest model (PostureProML), and (3) a Flutter-based mobile app (PostureProne). IMU data is sampled at 50Hz, processed via a Complementary Filter, and transmitted via Wi-Fi to a Supabase database for storage and FastAPI for inference.

Fig. 1

System architecture diagram showing IMU, ESP32, cloud, and app interactions

Hardware Design and Data Acquisition

The wearable system was designed to provide accurate, real-time lumbar posture monitoring during high-load strength training exercises. A 9-axis MPU-9250 inertial measurement unit (IMU)

was selected for its compact form factor and integrated accelerometer, gyroscope, and magnetometer capabilities. The sensor was positioned at the L4–L5 vertebral region (Fgure 2a, 2b), an anatomical landmark associated with maximum mechanical loading during Romanian Deadlifts (RDLs) (Fig. 2c), where compressive and shear forces can reach 18 kN and 3 kN, respectively[30, 31].

Fig. 2

(a) Participant wearing device over L4-L5 segment, (b) L4-L5 Spinal Segment [37], (c) Sensor positioning at the L4-L5 vertebral region.

The IMU was interfaced with an ESP32 microcontroller, chosen for its low-power wireless communication capabilities via Wi-Fi and integrated digital signal processing. The circuit was powered by a 3.7V Li-ion rechargeable battery with a voltage regulator, supporting continuous

operation for up to 4 hours. The hardware was compact and lightweight, designed for integration into athletic attire using an adjustable elastic strap that secured the device at the lumbar spine without restricting movement.

To reduce sensor drift and enhance signal fidelity, a complementary filter was applied for orientation estimation using the following equation:

!"#$%!&'()*+$, -.*%/0#.$1!"#$%!&'()*+$,-.*%/0#.$1

where θ is the filtered angle, ω is the gyroscopic angular velocity, a is the accelerometer-derived angle, Δt is the sampling interval, and α = 0.98 represents the filter gain.

A flat-surface zeroing protocol was implemented during calibration to ensure consistency in initial orientation readings. Sensor data were sampled at 50Hz and logged in CSV format for subsequent processing. Data acquisition was conducted with fourteen healthy adult participants (aged 18–25), each performing multiple repetitions of RDLs under guided supervision. A biomechanical expert annotated each repetition as “proper,” “rounded,” or “arched” based on visual assessment, forming the ground truth dataset for subsequent model training and evaluation.

Machine Learning Model: PostureProML

The PostureProML machine learning module was developed to classify lumbar posture during strength training using time-series data from the MPU-9250 IMU. Data were collected from 14 participants performing Romanian Deadlifts (RDLs) under supervised conditions, with postures—proper, rounded, and arched labelled by a biomechanical expert. Twelve time-domain statistical features were extracted from the raw IMU signals, including the mean, standard deviation, and peak values of the accelerometer (AccelX, AccelY, AccelZ) and estimated spine angle (AngleX,

AngleY). These features capture variations in body alignment and motion dynamics across repetitions. Feature selection and preprocessing were performed in Python using the Scikit-learn library.

A Random Forest classifier was chosen for its robustness against overfitting and suitability for low-dimensional feature spaces. Training was performed using both 70/30 and 80/20 train–test splits, with model generalizability confirmed via k-fold cross-validation. The final model achieved 94.5% classification accuracy on the test set. Model interpretability was enhanced using SHapley Additive exPlanations (SHAP), which identified spinal angle and linear acceleration features as the most influential predictors. To enhance clinical relevance, Health Risk Scores for arched/rounded postures were estimated based on frequency and duration.

Short-term: Moving averages, peak detection.

Medium-term: Autocorrelation, wavelet analysis.

Long-term: Trend decomposition and progressive change tracking.

These insights support adaptive feedback and risk-aware interventions.

For deployment, the trained model was serialized and hosted on a Flask-based REST API within a Supabase cloud backend, enabling real-time, and low-latency inference. This architecture ensures seamless integration with the wearable device and mobile application, providing scalable posture monitoring suitable for gym environments.

Mobile Feedback System: PostureProne App

The PostureProne mobile application was developed using the Flutter framework to provide real-time user feedback during strength training sessions. It interfaces with the wearable hardware via Wi-Fi and communicates with the cloud-hosted PostureProML model through HTTP requests, achieving a round-trip latency of less than 500ms (Fig. 3).

Fig. 3

Deployment architecture showing cloud processing pathways

The app delivers immediate audio-visual feedback to users based on posture classification outcomes. A color-coded gauge displays posture status (e.g., green for proper, red for rounded, yellow for arched) (Fig. 4), accompanied by optional haptic alerts to prompt real-time correction. Additional features include session tracking, posture history visualisation, and performance summaries to support user engagement and progress monitoring over time.

Usability testing was conducted with ten participants aged 18–25 following structured workout sessions. The app received an average rating of 4.2 out of 5 for intuitiveness and ease of use, with 90% of participants reporting that the feedback enhanced their posture awareness. These results support the app’s potential for deployment in fitness environments lacking direct supervision.

Fig. 4

Screenshot of PostureProne app showing color-coded posture gauge

System Integration and Testing

Comprehensive system-level testing was conducted to evaluate the performance, responsiveness, and user experience of the integrated wearable posture monitoring solution (Fig. 5). Key

evaluation metrics included classification accuracy, sensor drift, wireless data reliability, and user satisfaction.

Testing was carried out in a controlled gym environment with 14 participants performing Romanian Deadlifts (RDLs) while wearing the device. Real-time posture classification was assessed across three categories; proper, rounded, and arched, using the PostureProML model. The system demonstrated an overall classification accuracy of 94.5%, with stable wireless connectivity between the ESP32 microcontroller and the mobile application via Wi-Fi.

Sensor performance was analysed to assess drift and consistency. Angular deviation remained within a low range (1.8°–7.1°), validating the effectiveness of the complementary filter in reducing noise during dynamic movements. The end-to-end system achieved latency under 500 ms from data capture to feedback delivery, enabling timely corrective cues during exercise execution.

User satisfaction was measured using structured post-session surveys. Participants rated the intuitiveness, usefulness, and comfort of the system. The majority (90%) found the app interface intuitive, and 80% reported increased posture awareness during lifting tasks. These findings support the feasibility of deploying the system in real-world fitness settings where expert supervision may be limited.

Fig. 5

System flowchart

Ethical Considerations

All study procedures involving human participants were conducted in accordance with institutional ethical guidelines and the principles of the Declaration of Helsinki. Prior to participation, all individuals provided written informed consent after receiving a detailed explanation of the study's purpose, procedures, and potential risks.

To ensure data privacy and security, all sensor and classification data were anonymised and transmitted using Transport Layer Security (TLS) version 1.2 encryption protocols. Access to stored data was restricted to authorised personnel only. Participants were informed of their right to withdraw at any stage without consequence, and no identifying information was retained in the final dataset.

Results

Posture Data Collection and Labelling

Fourteen participants performed Romanian Deadlifts (RDLs) while wearing the lumbar-mounted IMU device. The MPU-9250 sensor captured tri-axial motion data at a sampling rate of 50Hz. Postural classification was based on three categories: proper, rounded, and arched, as determined by expert labelling.

One-way ANOVA tests revealed statistically significant differences in angular deviations across posture categories. Specifically, AngleX (Table 1, Fig. 6a) and AngleY (Table 2, Fig. 6b) measurements showed strong discrimination between postures:

AngleX: F(2, 39) = 68.20, p < 0.001

AngleY: F(2, 39) = 250.58, p < 0.001

Mean angular deviations were as follows:

Proper: 1.8° ± 0.45°

Rounded: 7.1° ± 0.65°

Arched: 6.4° ± 0.55°

These results validate the biomechanical sensitivity of the system in capturing posture-specific lumbar angles during high-load exercises.

Table 1
Tukey HSD Post Hoc Test Results for AngleX
Comparison	Mean Difference	P-Value	95%CI Lower	95%CI Upper	Significant
Archedvs Proper	4.7974	< 0.001	2.6606	6.9341	Yes
Archedvs Rounded	10.6425	< 0.001	8.5053	12.7797	Yes
Propervs Rounded	5.8451	< 0.001	3.6622	8.0280	Yes

Table 2
Tukey HSD Post Hoc Test Results for AngleY
Comparison	Mean Difference	P-Value	95% CI Lower	95% CI Upper	Significant
Arched vs Proper	0.0664	0.9796	-0.7383	0.8711	No
Arched vs Rounded	6.7818	< 0.001	5.9769	7.5867	Yes
Proper vs Rounded	6.7154	< 0.001	5.8933	7.5375	Yes

Fig. 6

(a)Sample plot of AngleX across three posture categories, (b)Sample plot of AngleY across three posture categories

Device Performance and Reliability

The wearable system demonstrated reliable hardware performance throughout all trials. Angular drift remained minimal across sessions, aided by the complementary filter and calibration routine. The ESP32 maintained stable Wi-Fi connectivity with the mobile application, and the 3.7V Li-ion battery provided up to 4 hours of continuous operation under active gym conditions without requiring recharge.

Machine Learning Model Performance

To evaluate the classification performance of the proposed PostureProML system, we compared its accuracy against two widely used baseline models—Support Vector Machine (SVM) and K-Nearest Neighbors (KNN). Additionally, we performed a SHAP (SHapley Additive exPlanations) analysis to determine the relative importance of input features in predicting posture categories. Figure 7 presents a comparative evaluation of the proposed PostureProML Random Forest classifier against baseline models. Figure 7(a) shows that PostureProML achieved the highest accuracy (94.5%) compared to the Support Vector Machine (92%) and K-Nearest Neighbors (90%), demonstrating superior classification performance. Inference latency remained < 500 ms, supporting real-time application (Fig. 7c). Figure 7 (b) illustrates the SHAP-based feature importance ranking, identifying lumbar angle (0.55) and vertical acceleration (0.35) as the most influential variables for posture classification. These findings confirm that the model not only outperforms competing approaches but also relies on biomechanically relevant features, enhancing its interpretability and applicability for real-time posture monitoring.

a b

Fig. 7

(a) Comparative accuracy of postureproml against baseline models (SVM, KNN), (b) feature importance derived from SHAP analysis, highlighting lumber angle and vertical acceleration as key discriminative metrics for posture classification. (c) Real-time processing latency.

User Experience and Qualitative Feedback

Usability testing was conducted with ten participants (aged 18–25) following hands-on use of the integrated system in a gym setting. The PostureProne mobile application received high user satisfaction ratings:

90% rated the app as intuitive

80% reported improved posture awareness during lifts

70% expressed willingness to adopt the system regularly

Participants praised the real-time audio-visual cues for enabling immediate posture corrections without disrupting workout flow.

Discussion

This study presents an interdisciplinary wearable system for real-time lumbar posture monitoring, designed specifically for gym-based strength training. The integration of low-cost hardware (MPU-9250 and ESP32), a machine learning classification model (PostureProML), and a Flutter-based mobile application (PostureProne) provides an accessible alternative to commercial posture monitoring systems, which are often expensive and lack exercise-specific feedback [3, 32].

Compared to previous IMU-based solutions designed for general movement monitoring[33], the proposed system offers improved contextual relevance by focusing on dynamic gym activities such as Romanian Deadlifts (RDLs). The high classification accuracy (94.5%) and sub-500ms latency support the feasibility of providing real-time feedback for injury prevention, an advancement over delayed or retrospective analysis models [34–36].

Hardware affordability further enhances system accessibility, with component costs approximately

$5–10 for the MPU-9250 and $4–6 for the ESP32 microcontroller, making it viable for personal or institutional deployment in fitness centres and training clinics. Additionally, the mobile app's intuitive interface and real-time alerts demonstrated strong user acceptance, with 90% of participants reporting improved posture awareness.

Nonetheless, several limitations must be addressed. The current study utilised a small sample size (n = 14), limiting statistical power and generalisability. Moreover, the testing was conducted in a controlled laboratory setting with a focus solely on RDLs. These constraints reduce the system’s applicability to a wider range of exercises or real-world conditions. Although the complementary filter effectively minimized sensor drift, dynamic multi-planar movements still pose challenges for inertial accuracy, an issue noted in prior work [34, 37].

Future work will address these limitations by expanding the participant pool to 30–50 individuals and incorporating additional compound movements such as squats and lunges. Secondly, expanding sensor placement and modalities may produce a more wholistic analysis on posture and injury prevention. Further, integrating magnetometer data and employing advanced sensor fusion algorithms (e.g., Madgwick or Kalman filters) will enhance 3D orientation tracking [38]. Longitudinal validation in real-world gym settings will also be pursued to assess long-term usability, data integrity, and behaviour change outcomes.

Conclusion

This paper presents a cost-effective, gym-oriented wearable system that successfully integrates IMU sensing, machine learning classification, and a mobile feedback interface for real-time lumbar posture monitoring. The PostureProML model achieved a 94.5% classification accuracy across key postural deviations, while the PostureProne app facilitated immediate corrective feedback with high user satisfaction. The system’s affordability, responsiveness, and usability position it as a practical tool for mitigating injury risks during unsupervised strength training, particularly in settings lacking access to expert guidance. Future developments will focus on expanding exercise coverage, enhancing multi-sensor fusion, and validating the system in

uncontrolled, real-world environments to support broader adoption and clinical impact in musculoskeletal health management.

References

Hamera, E., Goetz, J., Brown, C. & Van Sciver, A. Safety considerations when promoting exercise in individuals with serious mental illness, Psychiatry research, vol. 178, no. 1, pp. 220–222, (2010).

Hilmi, A. H., Abd Hamid, A. R. & Assyahid, W. A. R. Recent Advances in Ergonomic Posture Research: Assessing Innovations in Occupational Health and Musculoskeletal Disorder Prevention.

Panumasvivat, J., Surawattanasakul, V., Kwangsukstith, S., Mahakkanukrauh, C. & Kiratipaisarl, W. The relationship between working conditions, vehicle factors, and work-related musculoskeletal disorders affecting job satisfaction and job stress among motorcycle food delivery riders. Int. J. Hyg. Environ Health. 267, 114594 (2025).

Terry, M. A. & Alvarez-Vazquez, E. Efficacy of Wearable Devices in Monitoring and Treating Parkinson’s Disease Symptoms (A Systematic Review and Meta Analyses, IEEE Access, 2025).

f, Ananth & Rino, E. p. yogi-well: an ai-enabled wireless wearable for posture correction, yoga guidance & stress monitoring.

Kb, S. tech yoga: the smart mat revolution, i-Manager's. J. Embedded Syst., 13, 2, (2025).

Qin, W. et al. Wearable real-time multi-health parameter monitoring system. Int. J. Cloth. Sci. Technol. 36 (1), 17–33 (2024).

Hassanpour, A. & Yang, B. Contactless Vital Sign Monitoring: A Review Towards Multi-Modal Multi-Task Approaches, Sensors, vol. 25, no. 15, p. 4792, (2025).

Kapil, D., Wang, J., Olawade, D. B. & Vanderbloemen, L. AI-Assisted Physiotherapy for Patients with Non-Specific Low Back Pain: A Systematic Review and Meta-Analysis, Applied Sciences, vol. 15, no. 3, p. 1532, (2025).

10.

Heneweer, H., Staes, F., Aufdemkampe, G., van Rijn, M. & Vanhees, L. Physical activity and low back pain: a systematic review of recent literature, (in eng), Eur Spine J, vol. 20, no. 6, pp. 826 – 45, Jun (2011). 10.1007/s00586-010-1680-7

11.

Neamah, I., Shbeeb, H., Al-Shammary, A. & Al-Nidawi, R. Biomechanics of the Spine: A case study on the fourth and fifth lumbar vertebrae. J. Anthropol. Sport Phys. Educ. 8 (3), 23–27 (2024).

12.

Eynipour, A., Arjmand, N., Dianat, I., Soltanian, A. R. & Heidarimoghadam, R. Assessing musculoskeletal disorder risks in an automobile part manufacturing factory: a comparison study of biomechanical and ergonomic tools, Health Scope, 13, 2, p. e139610, (2024).

13.

Marcos-Pardo, P. J. et al. Improving spinal alignment through innovative resistance training with outdoor fitness equipment in middle-aged and older adults: a randomized controlled trial. Sci. Rep. 15 (1), 14499 (2025).

14.

Mehta, S. & Sarpal, S. S. Smart Monitoring of Leg Exercises in Fitness Centers: A CNN-LSTM Model for Enhanced Gym Safety and Efficiency, in 2024 2nd International Conference on Recent Trends in Microelectronics, Automation, Computing and Communications Systems (ICMACC), : IEEE, pp. 98–102. (2024).

15.

Lee, R. et al. Evidence for the effectiveness of feedback from wearable inertial sensors during work-related activities: A scoping review, Sensors, 21, 19, p. 6377, (2021).

16.

Paloschi, D. et al. Validation and assessment of a posture measurement system with magneto-inertial measurement units, Sensors, vol. 21, no. 19, p. 6610, (2021).

17.

Pourshahrokhi, N., Sun, Y. & Asadipour, A. Commercial and research-based wearable devices in spinal postural analysis: A systematic review, in EAI International Conference on IoT Technologies for HealthCare, : Springer, pp. 65–83. (2023).

18.

Wijegunawardana, I., Ranaweera, R. & Gopura, R. Lower extremity posture assistive wearable devices: A review. IEEE Trans. Human-Machine Syst. 53 (1), 98–112 (2022).

19.

Buisseret, F., Dierick, F. & Van der Perre, L. Wearable sensors applied in movement analysis Vol. 22, p. 8239 (ed: MDPI, 2022).

20.

Clark, B. K., Brakenridge, C. L. & Healy, G. N. The importance of research on occupational sedentary behaviour and activity right now Vol. 19, p. 15816 (ed: MDPI, 2022).

21.

Awolusi, I., Nnaji, C., Okpala, I. & Albert, A. Adaptation behavior of construction workers using wearable sensing devices for safety and health monitoring. J. Manag. Eng. 40 (1), 04023055 (2024).

22.

Kim, Y. J. Approach to Intelligent Garment for Posture Correction: A Preliminary Study for Integrating Aesthetics and Sensor Interaction. J. Vib. Eng. Technol. 13 (1), 60 (2025).

23.

Garcia-Jaen, M., Sebastia-Amat, S., Sanchis-Soler, G. & Cortell-Tormo, J. M. Lumbo-pelvic rhythm monitoring using wearable technology with sensory biofeedback: a systematic review, in Healthcare, vol. 12, no. 7: MDPI, p. 758. (2024).

24.

Lee, O. & Park, D. Associations Between Type-Specific Sedentary Behaviour and Low Back Pain: Evidence From a Large-Scale Cohort Study, Musculoskeletal Care, 23, 2, p. e70118, (2025).

25.

Hilmi, A. H., Hamid, A. R. A. & Ibrahim, W. A. R. A. W. Smart Wearables in Ergonomic Applications: Recent Advances and Challenges in Human-Machine Integration. Malaysian J. Ergon. (MJEr). 6, 24–38 (2024).

26.

Afeez, A., Abbey, O. & Jubril, A. Flexible and Stretchable Sensors for Health Monitoring in Smart Apparel, (2025).

27.

Traverso, S. Integration of IMU-based Motion Tracking Algorithms into Wearable Devices (for Human Joint Angle Estimation, Politecnico di Torino, 2023).

28.

Zhu, P. & Hu, Y. Carbon nanomaterials intelligent wearable devices for real-time athlete monitoring and performance tracking, Matéria (Rio de Janeiro), vol. 30, p. e20250327, (2025).

29.

Abhinav, V. et al. Advancements in Wearable and Implantable BioMEMS Devices: Transforming Healthcare Through Technology, Micromachines, vol. 16, no. 5, p. 522, (2025).

30.

Sciortino, V. Advanced Biomechanical Modeling of Spine. In-vitro Testing and Finite Element Modelling, (2024).

31.

Brost, S. Wearable Sensor-Based Estimation of Spinal Loading in the Sagittal Plane during Fluoroscopic Procedures (Queen's University (Canada), 2024).

32.

Salaorni, F., Bonardi, G., Schena, F., Tinazzi, M. & Gandolfi, M. Wearable devices for gait and posture monitoring via telemedicine in people with movement disorders and multiple sclerosis: a systematic review. Expert Rev. Med. Dev. 21 (1–2), 121–140 (2024).

33.

Villa, G. et al. Validation of a Commercially Available IMU-Based System Against an Optoelectronic System for Full-Body Motor Tasks, Sensors, 25, 12, p. 3736, (2025).

34.

Jenkins, L. & Weerasekera, R. Sport-related back injury prevention with a wearable device, Biosensors and Bioelectronics: X, vol. 11, p. 100202, (2022).

35.

Seçkin, A. Ç., Ateş, B. & Seçkin, M. Review on Wearable Technology in sports: Concepts, Challenges and opportunities, Applied sciences, vol. 13, no. 18, p. 10399, (2023).

36.

De Fazio, R., Mastronardi, V. M., De Vittorio, M. & Visconti, P. Wearable sensors and smart devices to monitor rehabilitation parameters and sports performance: an overview, Sensors, 23, 4, p. 1856, (2023).

37.

Alatise, M. B. Pose estimation and data fusion algorithms for an autonomous mobile robot based on vision and IMU in an indoor environment (University of Pretoria, 2021).

38.

Ratchatanantakit, N., Nonnarit, O., Sonchan, P., Adjouadi, M. & Barreto, A. A sensor fusion approach to MARG module orientation estimation for a real-time hand tracking application, Information Fusion, 90, pp. 298–315, (2023).

Data Availability

The data collected and analyzed during the current study are available from the corresponding author on reasonable request.

Ethics Statement

All experimental protocols were approved by a named institutional and/or licensing committee.

Funding

Declaration

This research received no external funding.

Conflict of Interest

The authors declare no conflict of interest.

Author Contribution

M.A. and J.Y. designed and conducted the study, M.A developed the wearable prototype, and performed data collection and J.Y worked on the Machine learning model and analysis. H.T and M. G supervised the project and finalised the manuscript. All authors reviewed and approved the final manuscript.

Yes