Iqbal, Mohammad and Supriyati, Endang (2012) Ekstraksi Ciri pada Pengenalan Sistem Isyarat Bahasa Indonesia Sensor Flex dan Accelerometer. In: Prosiding Seminar Nasional Embedded System: revitalisasi klaster industri perangkat telematika internasional. Pusat Penelitian Informatika Lembaga Ilmu Pengetahuan Indonesia, Bandung, pp. 1-8. ISBN 978-979-15035-1-8
![]() Preview |
PDF (Sampul)
prosiding-cover-dafisi.pdf - Published Version Download (249kB) |
![]() Preview |
PDF (Artikel)
prosiding-pp09-16.pdf - Published Version Download (398kB) |
Abstrak
Feature extraction is performed to obtain quantities that show the object specify to identity. The good feature extraction algorithm makes the classification process more effective and efficient. In this research, five types of feature extraction are developed, using statistical approach, quantization or combination of both. Sensors are used to make gloves, i. e. flex sensors to measure finger beding and accelerometer to measure finger bending and accelerometer to measure moement in the x, y, z axes. From the sensor data, the feature extraction is made. The obtained feature vector is used for sign recognition by applying Dynamic Time WarpingMethod (DTW) and Euclidean Distance. Reference data (template) that the best matches the dixtance measured by the most minimum value (distance). Tests carried out using a dataset with 1000 data consistes of 50 classes (word sign), where each class composed of 20 data. The test data using 10 data for each class, and the reference data using the rest i. e. 10 data for each class. The test results show that the achievd=ed highest accuracy are 99.6%.
Item Type: | Book Section |
---|---|
Kata Kunci: | sign language, flex sensor, accelerometer, ekstraksi isi |
Subjects: | Teknologi > Teknologi (umum) Teknologi > Teknik mesin dan mekanik Teknologi > T1 Teknologi (Umum) > Teknologi Informasi |
Program Studi: | Fakultas Teknik > S1 Teknik Informatika |
Depositing User: | Users 2 not found. |
Date Deposited: | 09 Jan 2013 03:01 |
Last Modified: | 22 Feb 2017 04:35 |
URI: | http://eprints.umk.ac.id/id/eprint/685 |
Actions (login required)
![]() |
View Item |