ArticlePDF Available

Human Fall Detection from Depth Images using Position and Velocity of Subject

Authors:
Yoosuf Nizam at Maldives National University
Yoosuf Nizam
Mohd Norzali Haji Mohd at Universiti Tun Hussein Onn Malaysia
Mohd Norzali Haji Mohd
Muhammad Mahadi Abdul Jamil at Universiti Tun Hussein Onn Malaysia
Muhammad Mahadi Abdul Jamil
Download full-text PDFDownload full-text PDF
Read full-text
Download citation

Abstract

Fall detection and notification systems play an important role in our daily life, since human fall is a major health concern for many communities in today's aging population. There are different approaches used in developing human fall detection systems for elderly and people with special needs such as disable. The three basic approaches include some sort of wearable, non-wearable ambient sensor and vision based systems. This paper proposes a human fall detection system based on the velocity and position of the subject, extracted from Microsoft Kinect Sensor. Initially the subject and floor plane are extracted and tracked frame by frame. The tracked joints of the subject are then used to measure the velocity with respect to the previous location. Fall detection is confirmed using the position of the subject to see if all the joints are on the floor after an abnormal velocity. From the experimental results obtained, our system was able to achieve an average accuracy of 93.94% with a sensitivity of 100% and specificity of 91.3%.
1-s2.0-S1877050917302090-ma
in.pdf
Content uploaded by Mohd Norzali Haji Mohd
Author content
All content in this area was uploaded by Mohd Norzali Haji Mohd on Mar 02, 2017
Content may be subject to copyright.
Human Fall Detection from Depth Images using Position and Velocity of Subje
ct.pdf
Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
... Although latency has been identified in the proposed activity monitoring of the elderly in the context of the healthcare industry, the sensors are adequately used for measuring acceleration forces. A depth sensor-based approach is proposed for monitoring and detecting the healthcare of elderly people [33]. The fall detection algorithm is used to detect the person through the robust Kinect sensors. ...
... However, latency has been identified in the proposed activity monitoring of the elderly in the context of the healthcare industry [52]. Nonetheless, the proposed sensor-based system does not accurately monitor the direction of movements [33]. Sometimes, technical issues occur in the system due to inaccurate sensing, but smart device performance is increased with the help of smart technologies using IoT [30,99]. ...
An architectural framework of elderly healthcare monitoring and tracking through wearable sensor technologies
Article
Full-text available
  • Jan 2024
  • MULTIMED TOOLS APPL
The growing elderly population in smart home environments necessitates increased remote medical support and frequent doctor visits. To address this need, wearable sensor technology plays a crucial role in designing effective healthcare systems for the elderly, facilitating human–machine interaction. However, wearable technology has not been implemented accurately in monitoring various vital healthcare parameters of elders because of inaccurate monitoring. In addition, healthcare providers encounter issues regarding the acceptability of healthcare parameter monitoring and secure data communication within the context of elderly care in smart home environments. Therefore, this research is dedicated to investigating the accuracy of wearable sensors in monitoring healthcare parameters and ensuring secure data transmission. An architectural framework is introduced, outlining the critical components of a comprehensive system, including Sensing, Data storage, and Data communication (SDD) for the monitoring process. These vital components highlight the system's functionality and introduce elements for monitoring and tracking various healthcare parameters through wearable sensors. The collected data is subsequently communicated to healthcare providers to enhance the well-being of elderly individuals. The SDD taxonomy guides the implementation of wearable sensor technology through environmental and body sensors. The proposed system demonstrates the accuracy enhancement of healthcare parameter monitoring and tracking through smart sensors. This study evaluates state-of-the-art articles on monitoring and tracking healthcare parameters through wearable sensors. In conclusion, this study underscores the importance of delineating the SSD taxonomy by classifying the system's major components, contributing to the analysis and resolution of existing challenges. It emphasizes the efficiency of remote monitoring techniques in enhancing healthcare services for the elderly in smart home environments.
View
... In healthcare, the detection of Parkinson's disease from video data has been done by Arifoglu et al. [13]. Fall detection and monitoring have been performed by Panahi et al. [14] and Nizam et al. [15] using video frames. In other applications, such as IoT-based video surveillance systems, the video data is analyzed for intruder detection and alarm system [16]. ...
How the ‘Things’ Speak: The Usage and Applications of Sensors in IoT
Chapter
Full-text available
  • Sep 2023
The Internet of Things (IoT) has been massively revolutionizing human lives for the last few decades. The powerful and steady advancements in the field of science and technology have aided this process immensely. As a result, almost all aspects of our lives have grown smarter. Nowadays, life is almost unthinkable in the absence of IoT-enabled smart devices, such as smart televisions, smart computers, smart-phones, smart fitness trackers, etc. Needless to say, all these devices enjoy ever-growing popularity in this era of smart technology. This development is propelled by the existing digital communication backbone – the Internet. The very Internet, over which human communication started just a few decades back, is now being used by each and everything in our surroundings, be it natural or man-made. The advent and inclusion of IoT in recent times have highlighted how trees, crops, fruits, chairs and tables, electrical appliances, and all other objects around us can interact with each other. They are capable of communicating as freely as humans, and based on such communications, these things’ can even behave smartly, individually, and in unison, by making informed decisions in real-time! Given that these things do not possess the gift of life naturally, their ability to express themselves comes from the use of numerous types of sensing devices, also called sensors. The intelligent manufacturing and easy availability of these miniature, cost-effective sensing devices have given a new shape to almost all aspects of our lives. The data regarding the behavior of things, as captured by sensors, is essentially what the things express, and it carries meaning in the particular setting. This data may then be processed and analyzed at the source, transmitted over the internet and processed in a cloud or remote machine. While in some day-to-day applications, the data is used directly for decision-making (for example, in smart electric appliances), in more critical problems, the data needs ample processing and analysis (healthcare, activity recognition, etc.). In the latter case, different mathematical model-based machine learning algorithms are utilized to learn hidden patterns or features in acquired data and extracted features. With the use of a trained learning algorithm called a classifier, the new data is then used for decision-making purposes. The choice of such algorithms is often dependent on the type of sensor data being used and the corresponding application area. Thus, it is seen that IoT-based systems find application in various domains, starting with research and development up to industry, agriculture, defense, etc. In reality, the progress of researchers in different domains leads to smart products that, in turn, make human lives easier. Research in several popular verticals, such as Human Activity Recognition, Remote Healthcare, Remote Monitoring, Smart Automation, Smart Agriculture, etc., has yielded many such products. This chapter focuses on the deep-seated relationship between IoT and sensors from the perspective of state-of-the-art research. It offers discussions on the usage of various types of sensing devices, associated data, and their contribution towards solving specific research problems in the respective IoT-based applications. This includes the Video Camera, Inertial Measurement Unit (IMU) Sensors, Ultrasonic Sensors, Electrocardiogram (ECG) Sensors, Passive Infra-Red (PIR) Sensors, Electromyogram (EMG) Sensors, and some commonly used sensing devices for Environmental and Agricultural Smart system development. A pertinent case study is also included to demonstrate the role of sensors in the development of IoT-based systems. This study also highlights how little effort it takes to implement an IoT-based data acquisition system. The different popular application areas are discussed thereafter in terms of some broad categories. This is followed by the description of some of the standard metrics used for evaluation and benchmarking the performance of smart sensing systems. The future of sensing devices has been discussed, followed by the pertinent challenges faced by IoT-enabled smart systems in implementation. Finally, the concluding remarks are offered. The chapter aims to serve as a wholesome source of knowledge, and relevant information to researchers and practitioners who wish to indulge in the development of smart IoT enables sensing systems.
View
... The most common machine learning algorithms that have been used to analyze data to ascertain whether a fall has occurred are k-Nearest Neighbors (k-NN), Support Vector Machines (SVM) and Deep Neural Networks (DNN) [13,23,25]. Support Vector Machines and k-Nearest Neighbor algorithms execute quickly and perform well in terms of accuracy. ...
An Analysis of Current Fall Detection Systems and the Role of Smart Devices and Machine Learning in Future Systems
Chapter
Full-text available
  • Mar 2023
Fall detection and prevention is a critical area of research especially as senior populations grow around the globe. This paper explores current fall detection systems, and proposes the integration of smart technology into existing fall detection systems via smartphones or smartwatches to provide a broad spectrum of opportunities for data collection and analysis. We created and evaluated three ML classifiers for fall detection, namely, k-NN, SVM, and DNN using an open-source fall dataset. The DNN performed the best with an accuracy of 92.591%. Recommendations are also included that illustrate the limitations of current systems, and suggest how new systems could be designed to improve the accuracy of detecting and preventing falls.
View
... Apabila hasil nya melebihi dari threshold yang sudah ditentukan, maka berarti ada yang terjatuh. Selain dari menghitung perubahan kecepatan dari setiap sumbu, ada juga penelitian yang melakukan perhitungan perubahan kecepatan dari titik-titik rangka pada rangka manusia [6]. Titik rangka utama yang digunakan dalam penelitian ini adalah titik rangka yang merepresentasikan kepala. ...
Sistem Deteksi Orang Jatuh Dengan Menggunakan Sensor Kamera Kinect Dengan Metode AdaBoost
Article
Full-text available
  • Oct 2021
Fall cases of elderly people aged 65 or above put their health at risk because it could lead to hip bone fracture, concussion, even death. Immediate help is needed if fall happened which is why an automatic and unobtrusive fall detection system is needed. There are three approaches in fall detection system; wearable, ambience, and vision-based. Wearable approach has the drawback of its obtrusive nature while ambience approach is prone to high false positive value. Vision-based approach is chosen because its unobtrusive nature and low false positive value. This study uses Kinect camera because of its ability on extracting skeletal data. The methods that are used in the fall detection system are AdaBoost method and joint velocity thresholding method. Thresholding method is used as a comparison to AdaBoost method. Both methods use skeletal data from the subject recorded by the Kinect camera. AdaBoost method compares the skeletal data with model that was made before while thresholding method compares the joint velocity value with the threshold value. System test is done using training data, test data, and real-time data. The average accuracy obtained from the system test with AdaBoost method is 91.75% and with thresholding method is 68.22%.
View
... There are many researches on feature extraction using deep learning algorithms. For example, researchers detected the falls after processing the acquired depth images [12], but the depth image resolution decreased as the distance of the target from the camera increased, and eventually it was difficult to complete background subtraction and segmentation of the depth images due to the low resolution. A method [13] combined multisensor, YOLOV3 and Lite Flow Net algorithms to detect falls, but the application conditions of the Lite Flow Net algorithm had certain limitations of light stability and small motion. ...
Fall detection based on OpenPose and MobileNetV2 network
Article
Full-text available
The proposed fall detection approach is aimed at building a support system for the elders. In this work, a method based on human pose estimation and lightweight neural network is used to detect falls. First, the OpenPose is used to extract human keypoints and label them in the images. After that, the modified MobileNetV2 network is used to detect falls by integrating both human keypoint information and pose information in the original images. The above operation can use the original image information to correct the deviation in the keypoint labeling process. Through experiments, the accuracy of the proposed method is 98.6% and 99.75% on the UR and Le2i datasets, which is higher than the listed comparison methods.
View
... meter for elderly fall detection. In this approach, if the acceleration exceeds a threshold value, it means that the person is in motion. At that moment, the depth sensor begins to extract information in order to detect a possible fall. However, the process requires calibrating the cameras and accelerometers, which increases its computational cost.Nizam et al. (2017) also used a Kinect sensor to study the speed and position of a person. Thus, if a high speed in a short time is detected, it is assumed that a fall has occurred. The fall is confirmed or discarded by analyzing the position of the body. This system has an average precision of 93,94 %.Mastorakis & Makris (2014) attempted elderly fall dete ...
Human Activity Recognition via Feature Extraction and Artificial Intelligence Techniques: A Review
Article
Full-text available
  • Sep 2022
Context: In recent years, the recognition of human activities has become an area of constant exploration in different fields. This article presents a literature review focused on the different types of human activities and information acquisition devices for the recognition of activities. It also delves into elderly fall detection via computer vision using feature extraction methods and artificial intelligence techniques. Methodology: This manuscript was elaborated following the criteria of the document review and analysis methodology (RAD), dividing the research process into the heuristics and hermeneutics of the information sources. Finally, 102 research works were referenced, which made it possible to provide information on current state of the recognition of human activities. Results: The analysis of the proposed techniques for the recognition of human activities shows the importance of efficient fall detection. Although it is true that, at present, positive results are obtained with the techniques described in this article, their study environments are controlled, which does not contribute to the real advancement of research. Conclusions: It would be of great impact to present the results of studies in environments similar to reality, which is why it is essential to focus research on the development of databases with real falls of adults or in uncontrolled environments.
View
Activity Classification and Fall Detection Using Monocular Depth and Motion Analysis
Article
Full-text available
  • Jan 2023
In the realm of senior care, falls remain the primary cause of hospitalization and mortality. Early detection can facilitate timely medical interventions and mitigate the associated risks. Hence, it is beneficial to set up a mechanism to identify and track such incidents. While wearable sensors are commonly used for this purpose, they may not be feasible for daily use. In contrast, surveillance cameras provide an unobtrusive and convenient monitoring solution. Despite the intricate nature of video frames, employing deep learning techniques for human recognition can enhance the performance of such systems. In our approach, activity classification is triggered only when a person is detected in the environment. We then process the extracted Global History of Motion (GHM) images with our Lightweight Deep Neural Network (LDNN). Additionally, we employ a dilated Convolution Long Short Term Memory (ConvLSTM) coupled with our LDNN to analyze the extracted sequence of monocular depth frames. Ultimately, the decision is made based on the combined information from both streams. Empirical results demonstrate the good performance of our fall detection system on the UP-Fall and UR datasets compared to the state-of-the-art. In terms of person classification, the Xception network achieved an F1-score of 98.46% on a sub-dataset derived from the COCO dataset, and 100% on the UP-Fall dataset.
View
View
A Review on Recent Developments on Detection of Fall
Article
  • Jun 2023
Remote patient monitoring has always been a difficult problem in the medical field. Fall detection during monitoring is essential because falls are unexpected behaviors that can seriously affect a person's health, particularly those who are older. Accidental falls have moved to the top of the lists of general health issues in the past few decades. A fall detection system, with the emerging development of the technology, aims to decrease the number of deaths, injuries and the economic burden on the healthcare system. This study presents an in-depth analysis of the latest published research on vision-based detection of falls. It also covers the merits, demerits, and challenges of the previous works of vision-based fall detection, and the future scope of the research is also summarized.
View
View
View
Classification of Human Fall from Activities of Daily Life using Joint Measurements
Article
Full-text available
  • Jan 2016
Falls are a major health concern to most of communities with aging population. There are different approaches used in developing fall detection system such as some sort of wearable, non-wearable ambient sensor and vision based systems. This paper proposes a fall detection system using Kinect for Windows to generate depth stream which is used to classify human fall from other activities of daily life. From the experimental results our system was able to achieve an average accuracy of 94.43% with a sensitivity of 94.44% and specificity of 68.18%. The results also showed that brutally sitting on floor has a higher acceleration, which is very close to the acceleration shown by fall. Even then the system was able to achieve a high accuracy in determining brutal movements with the use of joint positions, this is an indication that further improvements to the algorithm can make the system more robust.
View
Internal state measurement from facial stereo thermal and visible sensors through svm classification
Article
Full-text available
  • Jan 2015
Our main aim is to propose a vision-based measurement as an alternative to physiological measurement for recognizing mental stress. The development of this emotion recognition system involved three stages: experimental setup for vision and physiological sensing, facial feature extraction in visual-thermal domain, mental stress stimulus experiment and data analysis and classification based on Support Vector Machine. In this research, 3 vision-based measurement and 2 physiological measurement were implemented in the system. Vision based measurement in facial vision domain consists of eyes blinking and in facial thermal domain consists 3 ROI's temperature value and blood vessel volume at Supraorbital area. Two physiological measurement were done to measure the ground true value which is heart rate and salivary amylase level. We also propose a new calibration chessboard attach with fever plaster to locate calibration point in stereo view. A new method of integration of two different sensors for detecting facial feature in both thermal and visual is also presented by applying nostril mask, which allows one to find facial feature namely nose area in thermal and visual domain. Extraction of thermal-visual feature images was done by using SIFT feature detector and extractor to verify the method of using nostril mask. Based on the experiment conducted, 88.6% of correct matching was detected. In the eyes blinking experiment, almost 98% match was detected successfully for without glasses and 89\% with glasses. Graph cut algorithm was applied to remove unwanted ROI. The recognition rate of 3 ROI's was about 90%-96%. We also presented new method of automatic detection of blood vessel volume at Supraorbital monitored by LWIR camera. The recognition rate of correctly detected pixel was about 93%. An experiment to measure mental stress by using the proposed system based on Support Vector Machine classification had been proposed and conducted and showed promising results.
View
New Fast Fall Detection Method Based on Spatio-Temporal Context Tracking of Head by Using Depth Images
Article
Full-text available
  • Sep 2015
  • SENSORS-BASEL
In order to deal with the problem of projection occurring in fall detection with two-dimensional (2D) grey or color images, this paper proposed a robust fall detection method based on spatio-temporal context tracking over three-dimensional (3D) depth images that are captured by the Kinect sensor. In the pre-processing procedure, the parameters of the Single-Gauss-Model (SGM) are estimated and the coefficients of the floor plane equation are extracted from the background images. Once human subject appears in the scene, the silhouette is extracted by SGM and the foreground coefficient of ellipses is used to determine the head position. The dense spatio-temporal context (STC) algorithm is then applied to track the head position and the distance from the head to floor plane is calculated in every following frame of the depth image. When the distance is lower than an adaptive threshold, the centroid height of the human will be used as the second judgment criteria to decide whether a fall incident happened. Lastly, four groups of experiments with different falling directions are performed. Experimental results show that the proposed method can detect fall incidents that occurred in different orientations, and they only need a low computation complexity.
View
View
A Non-invasive Facial Visual-Infrared Stereo Vision Based Measurement as an Alternative for Physiological Measurement
Conference Paper
Full-text available
  • Nov 2014
Our main aim is to propose a vision-based measurement as an alternative to physiological measurement for recognizing mental stress. The development of this emotion recognition system involved three stages: experimental setup for vision and physiological sensing, facial feature extraction in visual-thermal domain, mental stress stimulus experiment and data analysis and classification based on Support Vector Machine. In this research, 3 vision-based measurement and 2 physiological measurement were implemented in the system. Vision based measurement in facial vision domain consists of eyes blinking and in facial thermal domain consists 3 ROI`s temperature value and blood vessel volume at Supraorbital area. Two physiological measurement were done to measure the ground truth value which is heart rate and salivary amylase level. We also propose a new calibration chessboard attach with fever plaster to locate calibration point in stereo view. A new method of integration of two different sensors for detecting facial feature in both thermal and visual is also presented by applying nostril mask, which allows one to find facial feature namely nose area in thermal and visual domain. Extraction of thermal-visual feature images was done by using SIFT feature detector and extractor to verify the method of using nostril mask. Based on the experiment conducted, 88.6\% of correct matching was detected. In the eyes blinking experiment, almost 98\% match was detected successfully for without glasses and 89\% with glasses. Graph cut algorithm was applied to remove unwanted ROI. The recognition rate of 3 ROI`s was about 90\%-96\%. We also presented new method of automatic detection of blood vessel volume at Supraorbital monitored by LWIR camera. The recognition rate of correctly detected pixel was about 93\%. An experiment to measure mental stress by using the proposed system based on Support Vector Machine classification had been proposed and conducted and showed promising results.
View
Human fall detection on embedded platform using depth maps and wireless accelerometer
Article
Full-text available
  • Oct 2014
  • COMPUT METH PROG BIO
Since falls are a major public health problem in an ageing society, there is considerable demand for low-cost fall detection systems. One of the main reasons for non-acceptance of the currently available solutions by seniors is that the fall detectors using only inertial sensors generate too much false alarms. This means that some daily activities are erroneously signaled as fall, which in turn leads to frustration of the users. In this paper we present how to design and implement a low-cost system for reliable fall detection with very low false alarm ratio. The detection of the fall is done on the basis of accelerometric data and depth maps. A tri-axial accelerometer is used to indicate the potential fall as well as to indicate whether the person is in motion. If the measured acceleration is higher than an assumed threshold value, the algorithm extracts the person, calculates the features and then executes the SVM-based classifier to authenticate the fall alarm. It is a 365/7/24 embedded system permitting unobtrusive fall detection as well as preserving privacy of the user.
View
Busuishengxue ranules mediate their effects upon non-severe aplastic anemia via mitogen-activated protein kinase/extracellular signal-regulated kinase pathway
Article
Full-text available
  • Feb 2014
  • J Tradit Chin Med
Objective To observe the clinical efficacy of Busuishengxue granules on non-severe aplastic anemia (NSAA) and investigate its effect on the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) pathway. Methods Sixty NSAA patients were divided equally into two groups. Subjects in the experimental group were treated with Busuishengxue granules, and the control group with Zaizaoshengxue tablets. The treatment course was 6 months and curative efficacy was compared between the two groups as well as with 10 healthy individuals. Flow cytometry (FCM) was used to detect the intracellular concentration of Ca2 + ([Ca2 +]i). Western blotting was employed to detect the expression of enzymes in the MAPK/ERK pathway. Results The efficacy of Busuishengxue granules was significantly better than that of Zaizaoshengxue tablets (P<0.05). Before treatment, expression of JNK, phospho-ERK 1/2 and p-JNK was higher, and [Ca2 +]i higher, than that of the control group (P< 0.05). After treatment with Busuishengxue granules, expression of all enzymes related to signal transduction pathways in the blood cells of NSSA patients were altered to different degrees. Conclusion Busuishengxue granules had a better effect with regard to improving symptom scores, increasing the number of blood leukocytes, and increasing hemoglobin levels than Zaizaoshengxue tablets, and they differed slightly in terms of increasing the number of platelets.
View
An Enhanced Fall Detection System for Elderly Person Monitoring using Consumer Home Networks
Article
Full-text available
  • Feb 2014
Various fall-detection solutions have been previously proposed to create a reliable surveillance system for elderly people with high requirements on accuracy, sensitivity and specificity. In this paper, an enhanced fall detection system is proposed for elderly person monitoring that is based on smart sensors worn on the body and operating through consumer home networks. With treble thresholds, accidental falls can be detected in the home healthcare environment. By utilizing information gathered from an accelerometer, cardiotachometer and smart sensors, the impacts of falls can be logged and distinguished from normal daily activities. The proposed system has been deployed in a prototype system as detailed in this paper. From a test group of 30 healthy participants, it was found that the proposed fall detection system can achieve a high detection accuracy of 97.5%, while the sensitivity and specificity are 96.8% and 98.1% respectively. Therefore, this system can reliably be developed and deployed into a consumer product for use as an elderly person monitoring device with high accuracy and a low false positive rate.
View
Fall Injuries in the Elderly
Article
  • Aug 1985
  • CLIN GERIATR MED
Injuries are the sixth leading cause of death in the 75-and-over population, with falls the leading cause of injury-related deaths. Hospitals and residential centers for the elderly have high rates of falls and injuries. With increasing age, patients in nursing homes have a corresponding increase in the proportion of fatal falls. The patterns of reduction of injury and mortality in the past two decades are likely to have been the result of our medical and trauma care system's impact on the outcome of less severe injuries. Major improvements in the future will come through prevention.
View