Perancangan Sistem Klasifikasi Suara Sirene Bencana Dari Early Warning System (EWS) Berbasis Internet Of Things (IoT) Untuk Penyandang Tunarungu

Pramono, Rizqy Ramadhan (2023) Perancangan Sistem Klasifikasi Suara Sirene Bencana Dari Early Warning System (EWS) Berbasis Internet Of Things (IoT) Untuk Penyandang Tunarungu. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07111840000153-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 April 2025. Download (3MB) | Request a copy

Abstract

Bencana adalah peristiwa yang dapat disebabkan oleh faktor yang mengakibatkan timbulnya kerusakan di sekitar. Early Warning System berfungsi untuk mendeteksi bencana dalam bentuk suara. Internet of Things dapat dimanfaatkan agar informasi dari Early Warning System dapat tersampaikan secara luas, terutama di smartphone dengan aplikasi Android. Namun, penyandang tunarungu tidak dapat mendengar suara. Maka dari itu dirancang sistem pendeteksi bencana berbasis visual kepada penyandang tunarungu menggunakan mikrofon, Raspberry Pi dan aplikasi Android. Untuk mendeteksi suara sirene dan alarm kebakaran, suara diubah menjadi gambar spectrogram yang kemudian dilatih menggunakan metode Convolutional Neural Network. Akurasi sebesar 0,9730 dan loss sebesar 0,0589. Peralatan yang disiapkan adalah Raspberry Pi, mikrofon, power supply, dan HP Android. Prinsip kerja alat, Raspberry Pi akan merekam suara dan setiap 3 detik diubah menjadi spectrogram dan diklasifikasi apakah suara tersebut sirene, alarm kebakaran, atau background noise. Jika terdeteksi sirene atau alarm kebakaran, data akan dikirim ke Firebase Realtime Database. Program di aplikasi Android berfungsi menerima data dari Firebase. Saat terdeteksi bahaya, aplikasi akan mengirim peringatan bahaya melalui notifikasi kepada seluruh pengguna aplikasi. Dilakukan dua pengujian menggunakan 30 suara sirene, 30 suara alarm kebakaran, dan 30 suara background noise. Pengujian pertama menguji akurasi sistem klasifikasi dalam mendeteksi suara. Sistem memiliki akurasi sebesar 93,34% untuk sirene, 96,67% untuk alarm kebakaran. Pengujian kedua menguji aplikasi dalam mengirim notifikasi tanda bahaya. Dibutuhkan rata-rata 6,901 detik untuk sirene, 4,257 detik untuk alarm kebakaran
======================================================================================================================================
Disaster is an event that can be caused by factors that result in damage around. Early Warning System functions to detect disasters in the form of sound. The Internet of Things can be utilized so that information from the Early Warning System can be widely conveyed, especially on smartphones with Android applications. However, deaf people cannot hear sounds. Therefore, a visual-based disaster detection system is designed for deaf people using microphones, Raspberry Pi and Android applications. To detect siren and fire alarm sounds, the sound is converted into spectrogram images which are then trained using the Convolutional Neural Network method. Accuracy is 0.9730 and loss is 0.0589. The equipment prepared is Raspberry Pi, microphone, power supply, and Android cellphone. The working principle of the tool, Raspberry Pi will record the sound and every 3 seconds it is converted into a spectrogram and classified whether the sound is a siren, fire alarm, or background noise. If a siren or fire alarm is detected, the data will be sent to the Firebase Realtime Database. The program in the Android app serves to receive data from Firebase. When a hazard is detected, the application will send a hazard warning via notification to all application users. Two tests were conducted using 30 siren sounds, 30 fire alarm sounds, and 30 background noise sounds. The first test tested the accuracy of the classification system in detecting sounds. The system has an accuracy of 93.34% for sirens, 96.67% for fire alarms. The second test tested the application in sending alert notifications. It takes an average of 6.901 seconds for sirens, 4.257 seconds for fire alarms

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Android, CNN, Early Warning System, Spectrogram, Tunarungu, Android, CNN, Deaf, Early Warning System, Spectrogram
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.774.A53 Android Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20201-(S1) Undergraduate Thesis
Depositing User:	Rizqy Ramadhan Pramono
Date Deposited:	07 Feb 2023 06:23
Last Modified:	07 Feb 2023 06:23
URI:	http://repository.its.ac.id/id/eprint/96352

Actions (login required)

View Item