Darindra, Anindya Maheswari (2025) Estimasi Kelelahan Otot Dan Kebugaran Kardiovaskular Dengan Electromyography (EMG) Dan Detak Jantung Berbasis Machine Learning Pada Latihan Sepeda Statis. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5023211025-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5023211025-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy |
Abstract
Pengelolaan kinerja fisik atlet merupakan aspek penting untuk meningkatkan performa dan mengurangi risiko cedera, overtraining, serta kelelahan yang berkepanjangan. Kelelahan adalah faktor utama yang dapat menghambat performa dan kesehatan fisik, terutama ketika intensitas latihan tidak disesuaikan dengan kondisi tubuh. Penelitian ini mengembangkan sistem pemantauan kinerja yang mengintegrasikan sinyal electromyography (EMG) dari otot quadriceps dan hamstring, data biomekanika yang didapatkan dari inertial measurement unit, serta data detak jantung yang diukur menggunakan photoplethysmography (PPG) pada jari tangan untuk mendeteksi tingkat kelelahan otot serta kebugaran kardiovaskular atlet selama latihan menggunakan sepeda statis. Algoritma yang digunakan memanfaatkan parameter fisiologis seperti perubahan amplitudo dan frekuensi sinyal EMG, data biomekanika, serta detak jantung, untuk mendeteksi kelelahan secara efektif. Analisis EMG dilakukan dengan mengekstraksi fitur seperti root mean square (RMS), integrated EMG (IEMG), simple square integral (SSI), mean absolute value (MAV), power spectrum (PS), mean frequency (MNF), dan median frequency (MDF). Sementara itu, sinyal PPG digunakan untuk mengukur variabilitas detak jantung selama latihan. Tingkat kelelahan tiap individu divalidasi menggunakan Borg Scale sebagai ground truth. Model Multilayer Perceptron Neural Network (MLPNN) dilatih menggunakan 80% data latih dan diuji pada 20% data uji, serta dievaluasi dengan metrik akurasi, precision, recall, dan f1-score. Hasil pelatihan menunjukkan akurasi tertinggi sebesar 89,96% pada kombinasi delapan fitur EMG terbaik. Sementara itu pada kombinasi enam fitur EMG dan dua fitur biomekanika menghasilkan akurasi hingga 89,25%. Temuan ini menunjukkan bahwa pemilihan fitur yang tepat memungkinkan deteksi kondisi kelelahan otot dengan akurasi tinggi dan generalisasi yang baik antar subjek dan sesi latihan.
=========================================================================================================================================
Athletes' physical performance management is crucial for enhancing performance and reducing the risks of injury, overtraining, and prolonged fatigue. Fatigue is a major factor that can hinder physical performance and health, especially when training intensity is not aligned with an individual’s physical condition. This study aims to develop a performance monitoring system that integrates electromyography (EMG) signals from the quadriceps and hamstring muscles, biomechanical data obtained from an inertial measurement unit (IMU), and heart rate data measured via fingertip photoplethysmography (PPG), to detect muscle fatigue levels and cardiovascular fitness during stationary cycling exercises. The algorithm utilizes physiological parameters such as changes in EMG signal amplitude and frequency, biomechanical data, and heart rate to effectively detect fatigue. EMG analysis involves extracting features such as root mean square (RMS), integrated EMG (IEMG), simple square integral (SSI), mean absolute value (MAV), power spectrum (PS), mean frequency (MNF), and median frequency (MDF). Meanwhile, the PPG signal is used to assess heart rate variability during exercise. Fatigue levels were validated using the Borg Scale as ground truth. A Multilayer Perceptron Neural Network (MLPNN) model was trained using 80% of the dataset and tested on the remaining 20%, evaluated with accuracy, precision, recall, and F1-score metrics. The training results showed a maximum accuracy of 89.96% using the eight best EMG feature combination. Additionally, combining six EMG features with two biomechanical features resulted in 89.25% accuracy. These findings indicate that proper feature selection enables accurate detection of muscle fatigue with strong generalization across subjects and training sessions.
Actions (login required)
 |
View Item |