Kontrol Pergerakan Kursi Roda Berbasis Head Gesture Menggunakan Long Short-Term Memory

Sari, Aiza Fuji (2025) Kontrol Pergerakan Kursi Roda Berbasis Head Gesture Menggunakan Long Short-Term Memory. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024211019-Undergraduate_Theses.pdf Restricted to Repository staff only Download (11MB) | Request a copy

Abstract

Mobilitas merupakan aspek penting bagi individu dengan keterbatasan fisik, seperti penderita tetraplegia yang mengalami kesulitan menggunakan kursi roda konvensional karena keterbatasan gerak pada tangan dan kaki. Penelitian ini mengembangkan metode kontrol kursi roda berbasis head gesture menggunakan kamera untuk mendeteksi gerakan kepala secara real-time.Teknologi MediaPipe digunakan untuk ekstraksi titik landmark wajah, sedangkan model Long Short-Term Memory (LSTM) diterapkan untuk mengklasifikasikan gesture kepala ke dalam lima kelas utama, yaitu Maju, Mundur, Kiri, Kanan, dan Stop. Dataset yang digunakan terdiri dari serangkaian citra berurutan dengan fitur landmark wajah yang telah dinormalisasi. Dua model LSTM diuji dan dianalisis performanya, dengan model pertama menunjukkan akurasi validasi hingga 100%. Pengujian dilakukan pada berbagai kondisi, termasuk variasi jarak kamera (50 cmdan70cm), intensitas pencahayaan (15 lux, 40 lux, dan 100 lux), serta dua perangkat komputasi (laptop dan Intel NUC). Hasil pengujian menunjukkan bahwa jarak kamera optimal adalah 50 cm dengan akurasi hingga 98%, serta intensitas pencahayaan 100 lux memberikan performa terbaik yaitu 97,33%. Perangkat NUC memiliki kecepatan pemrosesan (FPS) yang lebih tinggi dibandingkan laptop, sehingga lebih responsif. Metode ini juga mampu mengenali perintah dari pengguna baru dengan akurasi di atas 80%. Penelitian ini memberikan kontribusi pada pengembangan teknologi assistive yang memungkinkan penderita tetraplegia mengendalikan kursi roda secara mandiri melalui gerakan kepala tanpa menggunakan tangan.
=====================================================================================================================================
Mobility is a crucial aspect for individuals with physical limitations, such as those with tetraplegia who often face difficulties operating conventional wheelchairs due to restricted movement in both their hands and feet. This study proposes a head gesture-based wheelchair control method, utilizing a camera to detect head movements in real time. MediaPipe technology is employed to extract facial landmark points, while a Long Short-Term Memory (LSTM) model is applied to classify head gestures into five main categories: Forward, Backward, Left, Right, and Stop. The dataset consists of sequential image frames with normalized facial landmark features. Two LSTM models were developed and evaluated, with the first model achieving a validation accuracy of up to 100%. Testing was conducted under various conditions, including different camera distances (50 cm and 70 cm), lighting intensities (15 lux, 40 lux, and 100 lux), and computing devices (laptop and Intel NUC). The results indicate that a camera distance of 50 cm yields optimal performance with an accuracy of up to 98%, while 100 lux lighting intensity provides the best performance at 97.33%. The Intel NUC also demonstrated higher processing speed (FPS) than the laptop, resulting in better responsiveness. Additionally, the method successfully recognized commands from new users with an accuracy above 80%. This research contributes to the development of assistive technology that enables individuals with tetraplegia to independently control a wheelchair using head movements, without relying on hand operation.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Kursi roda, Head gesture, MediaPipe, LSTM, Wheelchair
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) R Medicine > R Medicine (General) > R856.2 Medical instruments and apparatus. T Technology > T Technology (General) > T59.7 Human-machine systems. T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition. Optical pattern recognition.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Aiza Fuji Sari
Date Deposited:	13 Jun 2025 08:46
Last Modified:	13 Jun 2025 08:46
URI:	http://repository.its.ac.id/id/eprint/119114

Actions (login required)

View Item