International Journal of Computer Science Trends and Technology (IJCST) – Volume 12 Issue 2, Mar - Apr 2024
RESEARCH ARTICLE OPEN ACCESS
Elderly Health Care and Virtual Interaction Platform Using
Machine Learning
Asst. Prof. Neethi Narayanan, Kalyani V Nair, Sreelakshmi A S,
Sreelekha A, Harsha T K
Assistant Professor, Department of Computer Science and Engineering, LBSITW, India
UG Scholar, Department of computer science and engineering, LBSITW, India
ABSTRACT
In the present digital era, social interaction and information sharing relies significantly on online platforms. However, elderly people
address major obstacles in this arena, particularly those who live alone. For this reason, this is a pressing issue that needs prompt
attention. Since many older persons lack the technological competence and access necessary to participate in online social
interactions, the primary issue is isolation which can lead to feelings of loneliness. Health monitoring is also a crucial aspect since
older adults often do not have access to prompt medical supervision, which can lead to health concerns going undiagnosed and
postponing treatment. Furthermore, crises can turn deadly, and elderly individuals living alone may find it challenging to get help in
the event of a medical emergency or injury. Our concept presents a platform for social interaction and elder care as a remedy to these
issues. This system proactively monitors living conditions and health of the elderly while enabling virtual social interactions with
loved ones. Through the use of technology and compassion, our effort aims to significantly improve the quality of life for senior
citizens. This will ensure their well-being in a digitally-driven culture where social interaction and health surveillance are essential.
With the goal to develop real-time monitoring and install an alert system that sends real-time alerts to caretakers in emergency
scenarios, we intend to simulate data reflecting real-time data such as health metrics (heart rate, temperature). Additionally, we
intend to develop a software-based platform that will allow individuals to communicate virtually through voice calls supported by
chat and hand gesture recognition features. Through this project, we intend to provide the elderly community with a healthy, user-
friendly platform that will enable them to live their best life with the help of cult technology.
Keywords: - “decision tree”, “machine learning”, “artificial intelligence”, “random forest”, “real-time sensor”.
I. INTRODUCTION
The landscape of machine learning has witnessed processing algorithms. Furthermore, machine learning models
remarkable strides in recent years, offering transformative can assess sensory inputs from wearable devices to monitor
potential across various domains, including healthcare and seniors' daily activities and health state, allowing for prompt
finance. Among these realms, elderly care emerges as a interventions and support as needed. However, while the
pivotal arena where the integration of machine learning prospective benefits of integrating machine learning in elderly
techniques holds meaningful promise, particularly in care are massive, numerous challenges and ethical
cultivating meaningful social interactions and enhancing the considerations like challenges of data privacy, algorithmic
well-being of older adults. As the count in number of elderly bias, psychological aspects and ethical considerations should
population across the globe increases, there is a growing need be dealt with smartly.
to find creative ways to deal with the challenges associated The incorporation of novel technologies into elderly care
with providing care for the elderly such as addressing the endeavours constitutes a revolutionary paradigm shift in how
widespread risk of social isolation and loneliness among we interact and serve elders. By utilizing the potential of
seniors, which poses major threats to their mental and physical artificial intelligence, we may develop more personalized,
health. Due to financial limitations and the difficulties immersive, and compassionate care experiences, improving
associated with respecting individual choices, conventional the quality of life for the elderly and fostering a culture of
care systems often fail to provide personal and stimulating healthy aging in our communities.
social connections.
Deep learning, a subset of machine learning is applied in a II. LITERATURE REVIEW
variety of ways in the field of elder care to improve social A remarkable contribution to the field of healthcare
interaction and overall quality of life. Conversational agents technology, particularly in addressing the specific needs of the
and chatbots that are adept at engaging elders in meaningful aging population especially concerned with fall detection and
conversations, memory rehabilitation sessions, or even virtual admission of rescue operations has been referenced in the
companionship can be powered by natural language paper [1]. The construction of an elderly healthcare system
ISSN: 2347-8578 www.ijcstjournal.org Page 103
International Journal of Computer Science Trends and Technology (IJCST) – Volume 12 Issue 2, Mar - Apr 2024
using high precision vibration sensors, a server, and a robotic exceptional image processing capabilities when compared to
companion is covered in this study. Experiments revealed that other algorithm categories. The research industry has
the system had an accuracy rate of 89% for identifying provided deep learning algorithms for creating intelligent
research participants' activities of the elderly. In order to systems for senior medical care attention. According to the
ensure that the system could recognize and alert help research, there are various deep learning algorithm
providers in the event that an elderly person lost applications in the creation of intelligent systems that help
consciousness due to a fall, a delay mechanism was added for with elderly healthcare. Deep learning was used for end-to-
the experimentation. The robot will make use of its end learning, as a classifier, and for feature extraction. The
microphone to ask the elderly person if they need any research also summarizes that several performance evaluation
assistance when it detects a fall. Additionally, the device will indicators had been considered for assessing the algorithms'
emit a warning if there isn't a human activity signal during the effectiveness. The model evaluation measures include true
next twenty seconds (perhaps due to elderly people becoming positive, true negative, false positive, false negative, recall, F1
unconscious after a fall). As compared to neural networks, the score, sensitivity, specificity and average absolute error along
experiment utilizes the noise-reduced data's autocorrelation with other metrics, even when majority of researchers focused
coefficient as the network's inputs. The outputs of the network on accuracy as their performance measure. In conclusion, the
remain the values of each state. The state with the highest paper urges the research community to embark on the
production value is actually used as the current estimated state. unexplored areas correlating deep learning and healthcare to
The sensor that was employed in this study was manufactured tackle hidden challenges hindering the development in this
by New Sensor Co., Ltd. to help with elderly monitoring on a area.
real time basis. The gadget is powered by a lithium battery, The focus of the paper [3] centres around the development of
which lasts for roughly six months before needing to be a prudent medical care system utilizing neural network
replaced. Its major advantage is that the range it can measure technologies. The authors delve into a relatively newer age
is sufficiently wide and also the privacy and ethical concept of “Virtual Nursing Home”. Even though the concept
consideration issues can also be avoided. has many potentials, especially in the domain of elderly care
The smart device-based robot is made up of servo actuators, and deals with providing remote inpatient care, it has many
micro control devices, robot bodies, and iOS smartphones. challenges that it has to tackle. For instance, older people
iOS devices are equipped with a multitude of sensors, such as require prompt assistance when they have physical discomfort
cameras, microphones, accelerometers, light sensors, or accidents. A more comprehensive service framework is
gyroscopes, touch screens, and compass. This facilitates required to address these issues. This research examines the
effective interaction between the elderly person and the robot issues with the conventional human posture recognition
assistant The paper also leaves scope for its future methods using artificial intelligence and allied technologies
developments such as usage of the sensor in a different Some of the basic networks are connected in order to increase
environment such as a bathroom and also the usage of hole convolution, widen the receptive field, and reduce
multiple sensors that help to determine the location of the parameters thereby lightening the network as a prerequisite to
elderly which helps to estimate their overall mobility rate and guaranteeing accuracy and speed. The paper concludes by
thereby predict the current physical condition of the elderly. highlighting that the content of the paper can aid a future flare
The paper [2] that was reviewed provides a thorough in the domain of Artificial Intelligence and in the times ahead,
examination of the correlation between integrating deep Artificial Intelligence can become entirely operational and
learning techniques with healthcare with a focus on the humanized.
comparison of the variety of existing systems. The paper The paper [4] suggests a social media interaction system that
included a number of case examples to help the reader may be employed to facilitate wholesome conversations
understand how feasible it is to use smart devices with deep between the younger generation and the elderly through
learning to help the elderly members of society with their already-existing social media. A computer system with a
medical needs. Deep learning algorithms have been utilized to microphone facility, speaker, camera facility and web access
address relevant issues in elderly healthcare services. Deep feature has been used to construct the suggested system. Older
learning algorithms are utilized for many applications, adults can use social media to voice retrieve and transmit
including hand gestures, facial recognition, depression, and information by simply pressing a button with their fingers
the detection of elderly falls. The research review reveals that thereby making it easily usable. The proposed model consists
the primary focus of the academic community is on detecting of Raspberry Pi, USB aided Microphone, Raspberry Pi
falls in elderly individuals. Most research projects focus on Camera, Speaker, Volume button, and Wireless LAN
identifying falls in the senior population. This is mainly since connector which facilities three operations: Receiving,
falls are a typical occurrence for senior people and carry a risk Sending and Automatic message receiving operations. When
of further health concerns like breaking a bone or developing the button during the receiving operation was pressed, it could
an injury in the body. Given that computer vision is a hear information from Google Calendar on average 0.74
promising area of study, it is not natural that CNN leads the seconds later. On average, it could also hear information from
charge in creating intelligent solutions to meet the elderly’s Twitter at 1.51 seconds. These are decent enough outcomes.
healthcare demands. CNN is widely recognized for its After capturing the audio and video, the transmission
ISSN: 2347-8578 www.ijcstjournal.org Page 104
International Journal of Computer Science Trends and Technology (IJCST) – Volume 12 Issue 2, Mar - Apr 2024
operation took, on average, 89.5 seconds. Since it was precise and real-time heart rate data, allowing developers to
straightforward to use and hear spoken messages, receiving capture and analyse pulse rate information to simulate diverse
operations had an excellent reputation. However, some scenarios. These real-world scenarios may encompass
unfavourable opinions on the mailing operation were received. monitoring heart rate during physical exertion, evaluating the
The primary cause was that, being a box type, the prototype impact of stress on cardiovascular health, or tracking sleep
system was novel to them. The paper concludes by patterns. Moreover, the MAX30100 sensor's versatility and
highlighting that it was successful in its attempt towards low power consumption enable its integration into wearable
building the system and that it shall work on focussing devices, facilitating continuous monitoring without hindering
towards the alteration of box type system to human type the user's mobility or comfort. Its ease of integration with a
system in its future works for improved performance. web interface to display the current reading makes
An unceasing monitoring facility for fall detection in older MAX30100 Pulse oximeter sensor emerge as a valuable tool
people is examined in paper [5]. The model makes use of the for real-time pulse rate collection in testing and development
open-source dataset from SisFall, which records elderly data environments. As technology continues to progress, the
using an accelerometer device. It also suggests that a potential for personalized and accessible health monitoring
wearable-sensor based fall detection system is the most solutions powered by sensors like the MAX30100 remains
efficient as it can detect the fall at any place and time. After promising, heralding a future where individuals can
computing a number of features, falls are identified using effortlessly monitor and optimize their health.
machine learning methods. To improve accuracy and
performance, two distinct machine learning algorithms, B. Continuous Temperature Detection Mechanism
Support Vector machine and decision trees which are utilized The DS18B20 temperature detection sensor is a reliable
and compared. In order to assess different trained models and solution for uninterrupted temperature sensing, in diverse
determine which algorithms are more efficient, several environments of the user’s body temperature. With its digital
parameters are computed, including the confusion matrix, output and one-wire communication protocol, the sensor
sensitivity, specificity, accuracy, training time, and prediction ensures seamless integration for developers aiming to monitor
time. To eliminate noise and unwanted glitches from sensor temperature fluctuations in real-time. By incorporating the
data, a 4th Order Butterworth filter with a 5Hz cutoff is used DS18B20 sensor into a web interface, users can conveniently
during data processing. For the data under consideration. access and visualize their current temperature values,
The decision tree algorithm performs better than the SVM enhancing accessibility and usability in temperature
algorithm. It is discovered that decision trees require less monitoring systems. The sensor's real-time functionality
computing time than SVM. While decision trees offer greater ensures users receive timely updates on their body
insight into the model's operation, SVM performs better right temperature, facilitating prompt interventions or adjustments
out of the box. Although decision trees are excellent for their as necessary. Whether monitoring temperature changes during
ease of use and interpretation, they are not very good at physical activity, tracking fever symptoms, or assessing
learning complex rules or extending to vast data sets. Due to environmental conditions, the sensor provides valuable
the reason that decision trees can properly categorize all the insights into user well-being. Its accuracy, reliability, and ease
attributes to its respective class, they provide greater precision of integration makes the DS18B20 sensor a preferred choice
than support vector machines (SVM). Decision trees had for real-time temperature sensing applications across medical
correctly detected falls with an accuracy of 96%. A fall settings, fitness trackers, and smart home devices.
detection system and simultaneous alert facility is also
included in the model by gathering real-time data. In the event C. Anomaly detection and Alert Mechanism
of fall of elderly, an alert is sounded in the form of call or Anomaly detection stands as a pivotal aspect of health
message to the caregiver to provide timely-care and monitoring systems, particularly concerning real-time data
intervention. The paper concludes by highlighting that a such as heart rate variability. The anomaly detection process
higher accuracy rate of detecting falls can be achieved by the commences by feeding real-time heart rate data into the model
model by considering larger datasets for training and also by fed by a wearable sensor, with machine learning algorithms,
identifying optimal features. such as random forest and decision tree, being trained on a
dataset containing normal heart rate patterns. When
III. METHODOLOGY confronted with new data, the model contrasts the input with
A. Real-time Pulse Rate Collection the reconstructed output and computes the magnitude of the
The MAX30100 Pulse oximeter sensor provides a robust reconstruction error. Significant deviations from anticipated or
solution for real-time pulse rate collection, offering direct normal behaviour in heart rate patterns manifest as anomalies.
measurement of an individual's heart rate without invasive The model evaluates the magnitude of the reconstruction
procedures. Its utilization of photoplethysmography (PPG) to error against a predetermined threshold or statistical measure
detect changes in blood volume in microvascular tissue makes derived from the trained dataset. Should the error surpass this
it ideal for continuous monitoring applications The threshold, it signifies the presence of an anomaly in the heart
MAX30100 sensor facilitates this process by delivering rate data. Upon anomaly detection, the system initiates an
alert mechanism to notify relevant stakeholders or caregivers.
ISSN: 2347-8578 www.ijcstjournal.org Page 105
International Journal of Computer Science Trends and Technology (IJCST) – Volume 12 Issue 2, Mar - Apr 2024
Integration options for alert delivery encompassing GSM detection, a pivotal aspect of the system, is achieved through
module alerts the caregiver through a call feature. Timely machine learning models. Notably, the Random Forest model
intervention in the event of reception of an alert proves demonstrates exceptional accuracy at 97.18%, followed
particularly crucial in scenarios where anomalies indicate closely by the Decision Tree model at 86.92%. These high
potential health risks or emergencies, such as irregular heart accuracy rates signify reliable anomaly detection capabilities,
rhythms or sudden changes in heart rate. Thus, anomaly crucial for early intervention in potential health issues.
detection in heart rate data, coupled with an efficient alert Temperature monitoring, facilitated by the DS18B20 sensor,
mechanism, bolsters the capacity of health monitoring systems adds another layer of comprehensive health tracking. Accurate
to detect and respond to abnormalities in real-time. temperature readings enable timely interventions in case of
abnormalities, which could indicate underlying health
D. Virtual Social Interaction concerns such as infections. Addressing the social and
The software-based virtual interface platform represents a emotional well-being of users, the system includes a virtual
novel approach to facilitating social interaction, particularly social interaction platform using hand gesture recognition
aimed at enhancing connectivity for elderly individuals. One which recognises the “peace” gesture and sends an email to
prominent feature is the e-mail facility aided by hand gesture the care giver to aid communication. By providing a platform
recognition, which serves as a vital communication tool for for social engagement, the system contributes to overall
elderly users to connect with peers, family members, and mental and emotional health. The system's effectiveness is
friends, thereby mitigating feelings of isolation and loneliness evident in its ability to offer comprehensive support,
often associated with aging. Central to the platform's user proactively managing health conditions and promoting social
interface is the integration of hand gesture recognition through connectivity. The high accuracy of anomaly detection
Image Processing modules. This innovative functionality underscores its reliability, while the combination of real-time
allows users to interact with the virtual interface seamlessly, monitoring and virtual support mechanisms contributes to
bypassing the need for traditional input devices such as improved quality of life for users, particularly the elderly
keyboards or touchscreens. The model recognizes the symbol population. Moving forward, future research could focus on
corresponding to “peace” gesture and generates an email to optimizing the system further, perhaps by refining machine
the preset caregiver with a fixed message “Your parent needs learning models or exploring additional features to enhance its
to contact you”. This makes routine tasks like making phone capabilities. Overall, the study highlights the potential of such
calls relatively easy for the elderly and allows them to contact integrated health monitoring systems to positively impact
their caregiver with minimal effort on their part. individuals' well-being by combining advanced technology
with personalized care and support.
E. Chatbot Assistance
The 24/7 availability chatbot stands as a technological V. CONCLUSION
beacon, tirelessly operating around the clock to deliver instant
assistance and support to users and caregivers. This feature is TABLE 1. MODEL PERFORMANCE COMPARISON
made possible through the implementation of keyword
recognition technology. The model is preset with some
answers for common questions concerning the elderly The combination of specialized social media platforms
healthcare domain. As soon as a question is asked, the model coupled with health monitoring technology and 24/7 chatbot
fetches and outputs its designated answer to the question as availability, provides an extensive solution to support the
response. A standout feature of this chatbot is its proficiency well-being of elders. By exploiting these, the elderly can
in automating repetitive tasks. Frequently Asked Questions preserve their independence and stay connected with loved
(FAQs) are addressed promptly and accurately, streamlining
the user experience and saving time for both users and
caregivers. Model Accuracy
Random Forest 97.18%
IV. RESULTS AND ANALYSIS Decision Tree 86.92%
The study presents a comprehensive health monitoring
system tailored to address the needs of individuals, Artificial Neural Network 23.33%
particularly the elderly, by integrating various advanced
technologies. It encompasses real-time pulse rate collection, ones by addressing health concerns in a pre-emptive manner.
temperature sensing, anomaly detection, virtual social A user-friendly interface tailored to the needs of elderly can
interaction, and chatbot assistance. Utilizing the MAX30100 be offered by social media platforms encouraging connections
Pulse oximeter sensor, the system effectively captures pulse and countering feelings of loneliness to promote
rate data crucial for assessing cardiovascular health. Anomaly psychological wellness among the elderly community. Health
ISSN: 2347-8578 www.ijcstjournal.org Page 106
International Journal of Computer Science Trends and Technology (IJCST) – Volume 12 Issue 2, Mar - Apr 2024
monitoring technology plays a crucial role in early detection [5] S. Badgujar, A.S. Pillai, “Fall Detection for Elderly
of health issues, enabling immediate action and reducing People using Machine Learning”2020 11th International
issues. Elderly can achieve timely medical attention and enjoy Conference on Computing, Communication and
better health outcomes with prompt actions supported through Networking Technologies (ICCCNT), Kharagpur, India
this technology. The accessibility of 24/7 chatbots guarantees
uninterrupted fellowship and assistance for senior citizens
particularly when human support may be restricted. These
chatbots can provide amity, provide answer to queries, and
provide recommendations or suggestions on various topics,
escalating the quality of life for the elderly. Thus, by
welcoming these ingenious solutions, the empowerment of the
elderly to lead independent lives with satisfaction while
staying healthy, and supported within their circle. A clear and
precise way towards addressing the unique needs of the aging
population in today's digital age can be represented by this
approach.
ACKNOWLEDGEMENT
In the first place, we extend our heartfelt gratitude to our
principal Dr. Jayamohan J for giving us all facilities and tools
for the development of this project. We record our sincere
gratitude and thanks to Dr. Anitha Kumari S, Department
Head of Computer Science and Engineering for encouraging
and supporting us in the development of this project. We use
this opportunity to extend our utmost gratitude to our project
guide Prof. Neethi Narayanan, Assistant Professor of
Computer Science and Engineering Department for her kind
and valuable guidance in completing our main project. Last
but not the least we thank all our friends and well- wishers for
their kind co-operation, immense support and useful
suggestions related to the project.
REFERENCES
[1] Shuai Shao, Jinsoek Woo, Kouhei Yamamoto, Naoyuki
Kubota. “Elderly Health Care System Based On High
Pecision Vibration Sensor.”, 2019 International
Conference On Machine Learning and
Cybernetics(ICMLC),Kobe, Japan.
[2] Mubarak Almutairi, Lubna A.Gabralla, Saidu Abubakar,
Haruna Chiroma, “Detecting Elderly Behaviours Based
on Deep Learning for Healthcare: Recent Advances,
Methods, Real-World Applications and Challenges”,
IEEE Access(Volume:10).
[3] Yue Geng, Chenlu Jia, Shuo Sun. “Research on
Intelligent elderly care system based on Neural
Network” 2022 International Conference on Big Data,
Information and Computer Network (BDICN). Shenyang
Institute of science and technology, Shenyang, 110000,
China.
[4] Toru. Kobayashi, Kazushige. Katsuragi. “Social Media
Mediation System for Elderly People” 2016 IEEE
International Conference on Consumer Electronics
(ICCE), Graduate School of Engineering, Nagasaki
University, Nagasaki, Japan.
ISSN: 2347-8578 www.ijcstjournal.org Page 107
