Yanuar, Rifky Dewani (2026) RANCANG BANGUN APLIKASI DETEKSI GERAKAN ANGKAT BEBAN MENGGUNAKAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK (CNN). S1 / D3 thesis, Universitas Kuningan.

[thumbnail of ABSTRAK] Text (ABSTRAK)
ABSTRAK.pdf

Download (883kB)
[thumbnail of BAB I] Text (BAB I)
BAB I.pdf

Download (138kB)
[thumbnail of BAB II] Text (BAB II)
BAB II.pdf
Restricted to Repository staff only

Download (959kB) | Request a copy
[thumbnail of BAB III] Text (BAB III)
BAB III.pdf
Restricted to Repository staff only

Download (1MB) | Request a copy
[thumbnail of BAB IV] Text (BAB IV)
BAB IV.pdf
Restricted to Repository staff only

Download (2MB) | Request a copy
[thumbnail of BAB V] Text (BAB V)
BAB V.pdf
Restricted to Repository staff only

Download (85kB) | Request a copy
[thumbnail of DAFTAR PUSTAKA] Text (DAFTAR PUSTAKA)
DAFTAR PUSTAKA.pdf

Download (216kB)
[thumbnail of LAMPIRAN] Text (LAMPIRAN)
LAMPIRAN.pdf
Restricted to Repository staff only

Download (5MB) | Request a copy
Official URL: https://rama.uniku.ac.id

Abstract

Latihan angkat beban memiliki risiko cedera signifikan apabila tidak dilakukan dengan teknik yang benar, terutama ketika pengawasan instruktur tidak dapat menjangkau seluruh pengguna secara bersamaan. Tanpa koreksi tepat, kesalahan teknik berisiko menimbulkan cedera muskuloskeletal serius. Penelitian ini bertujuan membangun aplikasi android untuk mengevaluasi kualitas teknik gerakan angkat beban secara otomatis menggunakan Convolutional Neural Network 1 Dimensi (CNN-1D) yang dikombinasikan dengan estimasi pose YOLOv8n-Pose secara real-time. Sistem dikembangkan melalui metode Prototype menggunakan 12 video beresolusi 1920×1080 piksel pada 30 FPS dengan total 144.611 frame, mencakup tiga kategori per gerakan pada gerakan Biceps Curl, Lateral Raise, Shoulder Press, dan Squat yang direkam di Doctor Fit Gym Kuningan. Data video diolah menjadi sequence time-series keypoints berdimensi (30, 51) sebagai masukan model CNN-1D untuk mengklasifikasikan kualitas teknik gerakan ke kategori benar atau salah. Sistem dilengkapi evaluasi rule-based berdasarkan parameter sudut sendi dan pola pergerakan tubuh untuk memberikan umpan balik rinci. Pengujian model CNN-1D menggunakan confusion matrix menunjukkan rata-rata akurasi di atas 99% dengan nilai precision, recall, dan F1-Score yang konsisten pada keempat gerakan. Pengujian UAT yang melibatkan 25 responden menghasilkan kelayakan 86,8% kategori sangat layak, mencerminkan penerimaan pengguna terhadap aplikasi sebagai solusi pendamping evaluasi teknik latihan mandiri bagi member gym pemula.

Weight training carries a significant risk of injury when not performed with proper technique, especially when instructor supervision cannot reach all users simultaneously. Without timely correction, technical errors risk causing serious musculoskeletal injuries. This study aims to develop an Android-based application to automatically evaluate the quality of weight training movement techniques using a One-Dimensional Convolutional Neural Network (CNN-1D) combined with YOLOv8n-Pose estimation in real-time. The system was developed using the Prototype method with 12 videos at 1920×1080 pixel resolution at 30 FPS totaling 144,611 frames, covering three categories per movement across Biceps Curl, Lateral Raise, Shoulder Press, and Squat recorded at Doctor Fit Gym Kuningan. The video data was processed into time-series keypoints sequences with dimensions (30, 51) as input to the CNN-1D model to classify movement technique quality into correct or incorrect categories. The system is equipped with rule-based evaluation of joint angles to provide corrective feedback. CNN-1D model testing using confusion matrices showed an average accuracy above 99% with consistent precision, recall, and F1-Score values across all four movements. UAT involving 25 respondents yielded a feasibility score of 86.8% in the very feasible category, reflecting user acceptance of the application among beginner gym members

Item Type: Thesis (S1 / D3)
Uncontrolled Keywords: Kata Kunci: Evaluasi Gerakan, Angkat Beban, Convolutional Neural Network, Pose Estimation, YOLOv8n-Pose
Subjects: T Technology > T Technology (General)
Divisions: Fakultas Ilmu Komputer > S1 Teknik Informatika
Depositing User: S. Kom Rifky Dewani Yanuar
Date Deposited: 07 Jul 2026 07:30
Last Modified: 07 Jul 2026 07:30
URI: https://rama.uniku.ac.id/id/eprint/5546

Actions (login required)

View Item
View Item