ESTIMASI POSISI PADA GEDUNG BERTINGKAT MENGGUNAKAN METODE FINGERPRINT BERDASARKAN DEEP NEURAL NETWORK
Total View This Week0
Total View This Week0
Institusion
Universitas Sriwijaya
Author
PRINITA AYUNINGTIAS (STUDENT ID : 09011381520061)
Reza Firsandaya Malik (LECTURER ID : 0025047609)
Firdaus Firdaus (LECTURER ID : 0221017801)
Reza Firsandaya Malik (LECTURER ID : 0025047609)
Firdaus Firdaus (LECTURER ID : 0221017801)
Subject
T1-995 Technology (General)
Datestamp
2020-01-14 02:24:14
Abstract :
In this study, a system of estimating positions with multi-building objects uses the RSS fingerprint method based on the Deep Neural Network (DNN). The position estimation system is carried out on a 3-story building located in the Faculty of Computer Science, Sriwijaya University with the floor research method. There are 3 trial scenarios with 2 scenarios using primary data and 1 scenario using secondary data namely RFKON. The fingerprint method consists of an offline phase and an online phase. Deep Neural Network architecture consists of an input layer, a hidden layer, and output layer. Accuracy results obtained from the three scenarios are in the first scenario of the first model 97.77%, the first experiment of the second model 98.27%, and the third experiment of the 93.45% third model. The results obtained in scenario 2 of the first model 98.14%, the second trial of the second model 99.07%, and the second trial of the third model 95.83%. The results obtained in the experiment scenario 3 were 99.23% in the RFKON_MB_WiFi comparison data database. In the second scenario, the reference point placement is reduced compared to the other two scenarios. This is due to the reduction in the location of adjacent reference points so that the interference impact is reduced.
In this study, a system of estimating positions with multi-building objects uses the RSS fingerprint method based on the Deep Neural Network (DNN). The position estimation system is carried out on a 3-story building located in the Faculty of Computer Science, Sriwijaya University with the floor research method. There are 3 trial scenarios with 2 scenarios using primary data and 1 scenario using secondary data namely RFKON. The fingerprint method consists of an offline phase and an online phase. Deep Neural Network architecture consists of an input layer, a hidden layer, and output layer. Accuracy results obtained from the three scenarios are in the first scenario of the first model 97.77%, the first experiment of the second model 98.27%, and the third experiment of the 93.45% third model. The results obtained in scenario 2 of the first model 98.14%, the second trial of the second model 99.07%, and the second trial of the third model 95.83%. The results obtained in the experiment scenario 3 were 99.23% in the RFKON_MB_WiFi comparison data database. In the second scenario, the reference point placement is reduced compared to the other two scenarios. This is due to the reduction in the location of adjacent reference points so that the interference impact is reduced.
File :
RAMA_56201_09011381520061_0025047609_0221017801_01_front_ref.pdf
RAMA_56201_09011381520061_0025047609_0221017801_02.pdf
RAMA_56201_09011381520061_0025047609_0221017801_03.pdf
RAMA_56201_09011381520061_0025047609_0221017801_04.pdf
RAMA_56201_09011381520061_0025047609_0221017801_05.pdf
RAMA_56201_09011381520061_0025047609_0221017801_06_ref.pdf
RAMA_56201_09011381520061_0025047609_0221017801_07_lamp.pdf
RAMA_56201_09011381520061_TURNITIN.pdf
RAMA_56201_09011381520061.pdf
Download
book
BibTex
Latex, Jabref
cloud_download
Original Resource
url resource Institution
