Sistem Identifikasi Manusia Bergerak Jatuh Berdasarkan Ekstraksi Suara dan Citra

Abstract

Falls are a major health problem around the world, especially in the world of healthcare because patient falls are the top worst problem that continues to occur. Most patients who fall out of bed are not witnessed. This is exacerbated by the various health problems that can result from falling. Remaining on the floor after a fall can cause trauma, serious injury and even death. Therefore, a fall detection system is needed so that people who fall can be given immediate help before they cause serious health problems. So in this study, we will create a fall identification system based on sound and image using the MFCC (Mel-Frequency Cepstrum Coefficients) method for sound extraction and LVQ (Learning Vector Quantization) for classification. Image processing using CNN (Convolutional Neural Network) method. In this system, both do not work together, but image processing works after sound processing. The system is able to detect falls with an overall accuracy of 93.3% for 15 times of sound and image processing tests.