Doctor of Philosophy (Ph.D.), University of Dayton, 2025, Electrical Engineering

The growing aging population requires innovative solutions in the healthcare industry. Telemedicine is one such innovation that can improve healthcare access and delivery to diverse and aging populations. It uses various sensors to facilitate remote monitoring of physiological measures of people, such as heart rate, oxygen saturation, blood glucose, and blood pressure. Similarly, it is capable of monitoring critical events, such as falls. The key challenges in telemonitoring are ensuring accurate remote monitoring of physical activity or falls by preserving privacy and avoiding excessive reliance on expensive and/or obtrusive devices. Our approach initially addressed the need for secure, portable, and low-cost solutions specifically for fall detection. Our proposed system integrates a low-power millimeter-wave (mmWave) sensor with a NVIDIA Jetson Nano system and uses machine learning to accurately and remotely detect falls. Our initial work focused on processing the mmWave sensor's output by using neural network models, mainly employing Doppler signatures and a Long Short-Term Memory (LSTM) architecture. The proposed system achieved 79% accuracy in detecting three classes of human activities. In addition to reasonable accuracy, the system protected privacy by not recording camera images, ensuring real-time fall detection and Human Activity Recognition (HAR) for both single and multiple individuals at the same time. Building on this foundation, we developed an advanced system to enhance accuracy and robustness in continuous monitoring of human activities. This enhanced system also utilized a mmWave radar sensor (IWR6843ISK-ODS) connected to a NVIDIA Jetson Nano board, and focused on improving the accuracy and robustness of the monitoring process. This integration facilitated effective data processing and inference at the edge, making it suitable for telemedicine systems in both residential and institutional settings. By developing a PointNet neural network for (open full item for complete abstract)

Committee: Vamsy Chodavarapu Ph.D. (Advisor); Kurt Jackson PT, Ph.D., GCS (Committee Co-Chair); Guru Subramanyam Ph.D. (Committee Member); Vijayan Asari Ph.D. (Committee Member) Subjects: Artificial Intelligence; Automotive Engineering; Biomedical Engineering; Computer Engineering; Electrical Engineering; Health Care Management; Robotics; Therapy

MS, University of Cincinnati, 2022, Engineering and Applied Science: Electrical Engineering

There are many research interests in human activity recognition, especially fall which is the major cause of serious injury for the elderly. Different technologies have been developed to detect falls, and fortunately, advancement in computer vision has attracted researchers to apply sophisticated systems for action recognition, posture estimation, and fall detection. The vision-based approach provides a non-invasive and reliable solution for fall detection among these various technologies. The overall goal of this thesis is to propose automatic human fall detection frameworks using confidence prediction and motion analysis. This thesis is composed of two pieces of work. The first work introduces a confidence-based fall detection system using multiple surveillance cameras. First, a model for predicting the confidence of fall detection on a single camera is constructed using a set of simple yet useful features. Then, the detection results from multiple cameras are fused based on their confidence levels. The proposed confidence prediction model can be easily implemented and integrated with single-camera fall detectors, and the proposed system improves the accuracy of fall detection through effective data fusion. Secondly, a flexible fall detection framework based on detecting a human object and analyzing the object's motion is proposed. Unlike many state of the art that require predefined thresholds to detect a fall, the proposed framework localizes and tracks a person in videos via object detection and motion analysis over a time window with appropriate length. As fall events may not look the same from different view angles, a multi-view fall dataset is used to train the proposed detection method. The framework is flexible for different use cases as it could incorporate unique object detection methods and work on videos captured from different angles. The proposed framework has produced promising detection results on several other datasets that outperfor (open full item for complete abstract)

Committee: Rui Dai Ph.D. (Committee Member); Xuefu Zhou Ph.D. (Committee Member); Wen-Ben Jone Ph.D. (Committee Member) Subjects: Electrical Engineering

Doctor of Philosophy, Case Western Reserve University, 2022, EECS - Electrical Engineering

Obstacle detection and warning can help elderly people enhance their mobility as well as their safety, especially in enclosed spaces (indoor environments). As people age, falling poses a significant risk, therefore providing mechanisms to prevent falls is vital to improve the safety and wellness of the elderly people population. Every year, millions of individuals in the United States are treated in emergency departments for fall-related injuries, which result in fractures, loss of independence, and even death. As a result, this issue must be addressed promptly. Fall prevention has been a focus of research for more than a decade, to enhance people's lives through the use of technology. This is primarily motivated by the impact that falls have in terms of mortality, morbidity, and social expense, which puts them on par with road traffic injuries in terms of mortality, morbidity, and social costs. Falls detection for elderly people can be essential to diminish the mortality rate and limit the associated health impacts. Technological solutions designed to automatically detect and inform a fall may be categorized into wearable and non-wearable solutions. Fall prevention systems take advantage of external sensors and wearable sensors where different motion characteristics are extracted from the collected data and are used to estimate the likelihood of a fall and alert the user in real-time. This work proposes an obstacle detection system to inhibit falls in the indoor environment. When obstacles are detected, the system will provide alarm messages to grab the user's attention. Because the elderly people spend a lot of their time at home, the proposed detection system is designed mainly for indoor applications. For this, firstly, obstacles are detected and localized, and then the information about the obstacles will be sent to the walker using an audio alert. In this dissertation, we present an assistive system for elderly people (open full item for complete abstract)

Committee: Dr. Kenneth A. Loparo (Committee Chair); Dr. Wyatt Newman (Committee Member); Dr. Farhad Kaffashi (Committee Member); Dr. Michael Fu (Committee Member) Subjects: Electrical Engineering

MS, University of Cincinnati, 2021, Engineering and Applied Science: Computer Science

According to a report produced by WHO in 2019, 1 in 11 people globally are considered elderly (over 65 years of age) accounting to more than 703 million elderly people. This calls for innovative prevention of quality livelihood in the elderly. The increase in older population results in careful monitoring of their daily activities. Fall related injuries and death has estimated about 30% of deaths in the USA. About 20-30% of elderly suffer from medium to severe injuries due to fall. So, monitoring and timely reporting the injuries is of utmost importance. Within this context, fall detection becomes a prominent field of study to cater better to the incapacitated elders. There are many devices and techniques commercially available to monitor their daily activities. Development in machine learning and deep learning has paved way for using artificial networks in fall detection. Video surveillance when combined with image processing has proven to be more effective than wearable devices. However, this will cost more memory space, computation, and money. Fall detection in still images is a conservative alternative but it is still under explored. Convolutional Neural Networks (CNN) are efficiently used in image processing. CNN can be used to get features from images. In this research work, VGG19 Network has been used to extract features from images. These feature maps are then used to classify the fall action using a KNN classifier. Dataset of depth images with a mixture of fall actions and daily activities has been used for this research purpose. This combination of classifiers has shown a result of approximately 60 % accuracy when compared to the individual classifiers.

Committee: Anca Ralescu Ph.D. (Committee Chair); Dan Ralescu (Committee Member); Kenneth Berman Ph.D. (Committee Member) Subjects: Computer Science

MS, University of Cincinnati, 2015, Engineering and Applied Science: Electrical Engineering

The objective of this thesis is to develop a low false-alarm-rate fall-down detection system for senior health care with an inertial measurement unit (IMU) and a microcontroller embedded with machine learning algorithm. Fatal delay in medical treatment caused by unconsciousness after seniors' fall-down results in thousands of death each year in US. Several different types of fall-down detection systems have been developed for seniors, especially those who live alone to send emergency notification to their families or emergency care agents. Recently, research has been conducted to reduce the high false-alarm-rate that fall-down detection systems suffer due to poor motion recognition. In this work, a prototype IMU-based smart fall-down detection system and low false-alarm-rate fall-down recognition algorithms have been newly developed. The fall-down detection system is composed of a combination of 3-axis gyroscope and a 3-axis accelerometer as the sensing unit, a push button and alarm as input/output, and a microcontroller as the control and data processing unit. The system combines angular velocity provided by the gyroscope and acceleration provided by the accelerometer to get the inclination data for orientation characterization. The impact data derived from the acceleration indicates the severity of the fall-down. The system has been calibrated by commercially available IMU to ensure its accuracy and stability. Fall-down and other motions are simulated with this system in this work. A low false-alarm-rate fall-down detection algorithm based on machine learning is developed and fully characterized in this work. The algorithm allows the detection system to be customizable by collecting false-alarm reports from the users. The algorithm extracts the pattern of similar false-alarm motions then trains the motion database to classify them as “safe motion”. Motions similar to known “safe motion” will be ignored by the detection system to reduce the false-alarm-rate (open full item for complete abstract)

Committee: Chong Ahn Ph.D. (Committee Chair); Wen-Ben Jone Ph.D. (Committee Member); Ian Papautsky Ph.D. (Committee Member) Subjects: Electrical Engineering

Master of Science in Computer Engineering (MSCE), Wright State University, 2011, Computer Engineering

Fall related injuries are the leading cause of death and hospitalization among the elderly. Falls among older people become a major problem facing hospitals and nursing homes. In this study we put an effort to design a wireless device capable of detecting falls with the hope that this study will provide a path towards better healthcare monitoring and better independent living for the elderly. In this project I showed how the fall detection device can be interfaced with different systems to achieve functionality without adding extra cost. For seniors who prefer to stay at their homes and live independently, the device can communicate with their smart phone to request help if needed. For hospitals and nursing homes, an infrared receiver and infrared signals decoding algorithms were implemented to interface with FastFind software to keep track of the location of the residents who fall or request help. There is also an option of having a live video feed from the specific room where the fall was detected.

Committee: Jack Jean PhD (Advisor); Yong Pei PhD (Committee Member); Meilin Liu PhD (Committee Member); Mateen Rizki PhD (Other); Andrew Hsu PhD (Other) Subjects: Engineering

Master of Science in Computer Engineering (MSCE), Wright State University, 2010, Computer Engineering

With the mortality rate of the elderly on the rise due to fall related incidents, fall detection has become an important entity in the geriatric health care sector. The study conducted in this thesis involved the design and development of a Wireless Data Acquisition System (WDAS) with Bluetooth capabilities interfaced to pulse width modulated gyroscopic sensors. Each sensor used has a reference output pulse width which changes when rotated about its sensitive axis.Part of the effort involves the design and development of two versions of printed circuit boards for the application. Each board includes a Parallax Propeller micro- controller in a TQFP (thin quad at pack) package, power regulator circuitry, and voltage shifting circuitry. The first version can interface with a single sensor and the second one has a three sensor interfacing capability. The software development effort includes designing the firmware for the microcontroller and developing a graphical user interface on Matlab to display and save sensor data sent through Bluetooth. The working of the system was tested with human subjects. The sensor outputs were obtained for induced falls in the coronal and the sagittal planes. Observations based on the subject testing data are summarized.

Committee: Jack S. N. Jean PhD (Committee Chair); Blair A. Rowley PhD (Committee Member); Yong Pei PhD (Committee Member); Gordana Gataric MD (Committee Member) Subjects: Engineering