Caryabudi, Samuel (2020) Sistem Deteksi Jatuh Low-cost dan Wearable dengan Koneksi Internet untuk Kalangan Lanjut Usia. Other thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
07311640000057-Undergraduate_Thesis.pdf Download (3MB) | Preview |
Abstract
Pada kalangan lanjut usia (lansia) umumnya akan terjadi penurunan kemampuan untuk menjaga keseimbangan sehingga membuat resiko untuk terjatuh ketika beraktivitas semakin meningkat. Ketika terjatuh, lansia dihadapkan dengan ancaman kondisi long-lie, yaitu keadaan terbaring di lantai dalam jangka waktu lebih dari satu jam setelah terjatuh karena keterlambatan pertolongan. Alat konvensional seperti Personal Emergency Response System (PERS) merupakan solusi yang kurang memuaskan untuk memanggil pertolongan karena banyak faktor-faktor yang menyebabkan alat tersebut tidak diaktifkan pengguna. Telah dikembangkan berbagai metode dan algoritma deteksi jatuh otomatis untuk mengatasi kekurangan tersebut, namun umumnya penelitian-penelitian tersebut kurang menjelaskan alur pemberian notifikasinya secara detail atau metode komunikasi yang dipilih kurang optimal. Pada sistem deteksi jatuh yang sudah tersedia secara komersial, terdapat beberapa kendala seperti rentan untuk terjadi kesalahan deteksi, fitur minimal, dan stasiun komunikasi khusus yang tidak portabel dan jangkauannya terbatas. Pada tugas akhir ini diusung suatu perwujudan sistem nyata dari algoritma deteksi jatuh berbasis threshold dengan memanfaatkan tiga keluaran tri-axial accelerometer yang diproses dengan Sum Vector Magnitude (SVM). Sistem dilengkapi dengan sistem notifikasi berupa bunyi secara langsung dan notifikasi yang dikirimkan ke smartphone melalui koneksi internet dengan protokol MQTT. Dilakukan simulasi kondisi jatuh realistis oleh subjek normal bukan lansia untuk mendapatkan threshold yang dipakai pada sistem, kemudian dilakukan pula simulasi beberapa contoh activities of daily living (ADL) lansia untuk mengetahui tingkat keberhasilan sistem dalam membedakan kondisi jatuh dan bukan jatuh. Sistem yang diwujudkan dapat membedakan kondisi jatuh dan bukan jatuh dengan tingkat keberhasilan tinggi, yaitu dengan persentase akurasi 100% pada posisi sensor di bagian dada dan 99,04% pada posisi sensor di bagian punggung, serta berhasil memberikan notifikasi ketika terdeteksi jatuh pada penggunanya. Dengan koneksi internet yang telah berhasil dibangun dalam penelitian ini, maka penambahan berbagai fitur pada sistem ini sangat memungkinkan.
==========================================================
Elderly people often experience a decline on the ability to maintain balance, so the risk of fall when performing activities increases. When elderly falls, they are faced with the threat of a “long-lie” condition, which is a condition/state of lying on the floor for more than one hour after falls because of delayed help. Conventional solution to call for help such as the Personal Emergency Response System (PERS) is unsatisfactory due to many factors that cause the user to not activate the device. Various automatic fall detection algorithms and methods have been developed to overcome that drawback, but those research usually did not describe the detailed notification flow or the communication method used was not optimal. There are several problems on fall detection systems that are commercially available such as vulnerability to false detection, minimal features, and the use of special communication station that is not portable and has limited range. This final project presents a fall detection system that uses a threshold-based fall detection algorithm that utilized the three outputs from tri-axial accelerometer which are then processed by Sum Vector Magnitude (SVM) method. This system is equipped with a notification system that provides alerting sound and notifications sent to smartphones via internet connection using MQTT protocol. A realistic simulation of fall was conducted by normal young subjects to obtain a threshold that is used on the system, then a simulation of some examples of elderly activities of daily living (ADL) was also conducted to determine the success rate of the system in distinguishing fall and non-fall conditions. The realized system is able to distinguish fall and non-fall conditions with a high success rate, which is 100% accuracy percentage for sensor position on the chest and 99.04% for sensor position on the back, and successfully provide necessary notifications when a fall is detected on the user. With the internet connection that has been successfully built in this study, it is possible to add various features to this system.
| Item Type: | Thesis (Other) |
|---|---|
| Uncontrolled Keywords: | internet, jatuh, lansia, long-lie, notifikasi, sistem deteksi jatuh, internet, fall, elderly, long-lie, notification, fall detection system |
| Subjects: | T Technology > T Technology (General) T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK3070 Automatic control T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5102.9 Signal processing. |
| Divisions: | Faculty of Electrical Technology > Biomedical Engineering > 11410-(S1) Undergraduate Thesis |
| Depositing User: | Samuel Caryabudi |
| Date Deposited: | 26 Aug 2020 06:22 |
| Last Modified: | 31 Aug 2023 07:26 |
| URI: | http://repository.its.ac.id/id/eprint/80857 |
Actions (login required)
 |
View Item |