Pengenalan Bahasa Isyarat Tangan Menggunakan Depth Image

Pajar, Try Yuliandre (2018) Pengenalan Bahasa Isyarat Tangan Menggunakan Depth Image. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

[img]
Preview
Text
07111340000141_Undergraduate_Theses.pdf - Accepted Version

Download (5MB) | Preview

Abstract

Sistem pengenalan bahasa isyarat menggunakan pengolahan visual digital. Pada tugas akhir ini sistem mengambil citra gambar menggunakan depth sensor. Depth sensor digunakan untuk mendapatkan gambar tangan sehingga tahap pengambilan kontur berbeda dibandingkan dengan kamera RGB. Depth sensor yang diatur jarak pembacaanya dapat menghasilkan gambar kontur tangan. Menggunakan metode convex hull, convexity defects, dan pusat massa gambar dapat menghasilkan nilai-nilai yang dapat dilatih untuk melakukan pengenalan pada tahap uji cobanya. Sistem ini dapat menangkap citra tangan dari jarak 50cm hingga 65cm. Sistem ini dilatih menggunakan artificial neural network dengan dua kondisi percobaan. Percobaan pertama menggunakan delapan output berdasarkan koordinat yang didapat. Percobaan kedua menggunakan tiga input berdasarkan panjang garis dan luas. Hasil yang dicapai sistem ini yaitu dapat mengenali gestur bahasa isyarat tangan berdasarkan hasil pelatihan. Hasil pelatihan ditentukan dari elemen penyusun neural network dan banyaknya iterasi yang dilakukan, pada ragam huruf yang sedikit akurasi hasil pelatihan dapat memenuhi target output sebaliknya jika ragam huruf bertambah banyak. ==================================================================================================== The sign language recognition system uses digital visual processing. In this final project the system takes image image using depth sensor. Depth sensors are used to get hand drawings so that the contour capture stage is different than the RGB camera. Depth sensors that can be adjusted the distance of the reader can produce hand contour images. Using convex hull, convexity defects, and image center methods can generate trained values for introduction to the test phase. This system can capture hand images from 50cm to 65cm. The system is trained using an artificial neural network with two experimental conditions. The first experiment uses eight outputs based on the acquired coordinates. The second experiment uses three inputs based on line length and area. The result of this system is to recognize gesture of hand sign language based on training result. The results of the training are determined by the neural network's constituent elements and the number of iterations performed, on the slightly different letters the accuracy of the training results can meet the opposite output targets if the variety of letters increases.

Item Type: Thesis (Undergraduate)
Additional Information: RSE 006.42 Paj p-1 3100018074566
Uncontrolled Keywords: Hand Sign Language, Kinect, Machine Vision, depth sensor.
Subjects: T Technology > T Technology (General)
T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques
Divisions: Faculty of Industrial Technology > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User: Try Yuliandre Pajar
Date Deposited: 22 Mar 2020 01:47
Last Modified: 22 Mar 2020 01:47
URI: http://repository.its.ac.id/id/eprint/50002

Actions (login required)

View Item View Item