International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 09 Issue: 06 | June 2022
p-ISSN: 2395-0072
www.irjet.net
A Review Paper on Elderly Fall Detection Gouri Nandhana1, Hema S2 1PG
Student, Dept. of Electronics & Communication Engineering, LBSITW, Kerala, India Professor, Dept. of Electronics & Communication Engineering, LBSITW, Kerala, India ---------------------------------------------------------------------***--------------------------------------------------------------------2Assistant
Abstract - According to World Health Organization (WHO),
optimization algorithm (SSOA) with variational autoencoder (VAE) based classifier is used for the classification of fall and non-fall events. In case of a fall event, an alert is given to the caretakers via smartphone. The experiment highlighted performance with the maximum accuracy of 99.57%.
falls are the second driving cause of accidental deaths around the world after road accidents. For elderly adults, fall could be highly risky and might cause life threatening health issues. It can be critical if the injured person does not get quick assistance. Due to the ever growing population of elderly people, there is a dramatic increase in fall detection. Therefore a fall detection system is used to detect a fall and to provide fast assistance for the person who is prone to fall. Currently multiple ideas exist to prevent the elderly from falling by means of technology. The main purpose of this review is to highlight some of the previous studies used for elderly fall detection.
Deok-Won Lee et al. proposed the double check method using the Inertial Measurement Unit (IMU) sensor and mobile robot [2]. A subject wearing the IMU sensor randomly repeats several actions like falling, standing, sitting and walking. The collected data sets are then input to the trained Recurrent Neural Network (RNN) based fall detection model and results are monitored. The IMU sensor is used to continuously track the user’s movement in real time. If a fall is detected, the robot moves to the corresponding area using the location information provided by the IMU sensor. The robot then acquires images using RGB sensors mounted on it. These RGB images are the input to the CNN algorithm and thus it double checks whether a fall truly occurred. This method gives 100% accuracy. The proposed method increases the cost and time of falls but could minimize the occurrence of false alarms and maximize the fall detection precision.
Key Words: Fall detection, Elderly care, Elderly Population, Injury.
1. INTRODUCTION Fall is an unplanned descent to the ground with or without injury. Fall accidents form one of the most important health problems in the ageing population. Lack of balance and fall might be symptoms of serious health issues. In this paper Elderly people are mainly focused because they are groups of people who are prone to illness and are not capable of protecting themselves and are most probably left unaccompanied at homes. Falls can cause physical injury and mental trauma which can even lead to anxiety and depression in elderly people. Nevertheless of the cause for a fall, it can be critical if the injured person does not get quick assistance. In recent years, the declining birth rate and aging population have gradually brought countries into an ageing society. Especially in India elderly population will increase to 12% of the national population by 2025 with 8%-10% requiring utmost care. Hence a fall detection system is an important component for elderly care. Fall detection system is needed to detect a fall and to provide fast assistance to the elderly person.
X.Cai ei al. proposed a multichannel convolutional fusion dense block strategy for fall detection [3]. The dense block strategy is used to obtain rich information with its densely connected layers and can compress network with less computation and fewer parameters, which will be beneficial for the fall detection. The process of this method is divided into testing and training phase. Ten consecutive frames are input into both phases to make use of spatio-temporal information. The feature map contains both spatial and temporal information obtained from input frames. To train the model supervised learning is applied. During testing phase classification result can be obtained. The proposed method gives 96.6% accuracy. Yen-Hung Liu ei al. proposed a pose estimation based fall detection algorithm using RGB camera [4]. The data set used is highly imbalanced, which meets real-world situation. OpenPose is used to extract the skeleton information from the images. Then feature extraction and feature scaling is done to help the model learn more effectively. For classification, machine learning approach is used. An alarm will sound when the classification model detects a fall event. This method gives 94.2% accuracy.
2. FALL DETECTION METHODS Thavavel Vaiyapuri et al. proposed An Internet of Things (IoT) enabled elderly fall detection model using optimal deep convolutional neural network [1]. Firstly, an IoT device captures the input video. The input video is then preprocessed in three levels like resizing, augmentation, and min-max based normalization. Then for feature extraction Squeeze Net model is employed to extract useful feature vectors for fall detection. Finally, a Sparrow search
© 2022, IRJET
|
Impact Factor value: 7.529
Chalavadi Vishnu ei al. proposed a fall motion mixture model (FMMM) approach for human fall detection representing fall
|
ISO 9001:2008 Certified Journal
|
Page 263