Kontrol Pergerakan Kursi Roda Berbasis Lip Gesture Menggunakan CNN

Putra, Dimas Triananda Murti (2024) Kontrol Pergerakan Kursi Roda Berbasis Lip Gesture Menggunakan CNN. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5024201012_Undergradute_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (12MB) | Request a copy

Abstract

Seseorang penyandang disabilitas seperti quadriplegia membatasi mobilisasi sehari-hari, memerlukan alat bantu seperti kursi roda. Namun agar melakukan mobilisasi tersebut secara mandiri dibutuhkan kursi roda elektrik yang dikendalikan melalui gerakan bibir, terintegrasi dengan teknologi pengenalan gestur menggunakan mediapipe dengan model CNN. Model ini optimal dalam mendeteksi pose bibir ketika menggunakan 10 lapis konvolusi dengan lapisan 32, 64, dan diakhiri dengan lapisan dense 128 dan terdapat dropout di tiap lapisan konvolusinya, menunjukkan bahwa pencahayaan 120 lux dan jarak antara bibir dengan kamera 70cm hingga 110cm meningkatkan akurasi model tersebut. Pengujian menunjukkan NUC lebih unggul darilaptop dalam kecepatan FPS karena perbedaan generasi CPU dan besar RAM. Delay rata-rata antar pengiriman data ke motor adalah 0.327 detik, dengan waktu inferensi rata-rata 0.08575 detik. Rata-rata kestabilan gerak motor kursi roda untuk pendeteksian selama 10 detik adalah 10.703.
=====================================================================================================
People with disabilities like quadriplegia face daily mobility restrictions and often require aids such as wheelchairs. For independent mobility, electric wheelchairs controlled through lip movements are essential, integrated with gesture recognition technology using mediapipe and a CNNmodel. This model is optimized for detecting lip poses using 10 convolutional layers with depths of 32 and 64, ending in a dense layer of 128, and includes dropout at each convolution layer, showing that illumination of 120 lux and a distance between the lips and camera of 70 to 110cm enhances model accuracy. Tests show that NUC outperforms laptops in FPS speed due to CPU generation differences and larger RAM. The average delay in data transmission to the motor is 0.327 seconds, with an average inference time of 0.08575 seconds. The average stability of the wheelchair motor movement for detection over 10 seconds is 10.703.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Mediapipe, Kursi Roda, Bibir Mediapipe, Wheelchair, Lip
Subjects:	T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques 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:	Dimas Triananda Murti Putra
Date Deposited:	31 Jul 2024 12:49
Last Modified:	31 Jul 2024 12:49
URI:	http://repository.its.ac.id/id/eprint/111318

Actions (login required)

View Item