2. Introduction
ONE of the most interesting features in modern portable and wearable devices is Human
Activity Recognition (HAR). It refers to the ability of a system to identify the activities
executed by a person, by processing data acquired from a set of sensors, which monitor
movements of parts of the human body . Both image and inertial sensors are used to build
HAR systems but the evolution of MEMS has largely contributed to make inertial-based HAR
systems more cost-efficient than the image-based counterpart . Therefore, HAR systems, based
on accelerometers, gyroscopes or heterogeneous Inertial Measurement Units (IMUs), are
becoming very popular in a number of handheld, embedded, and wearable devices . They are
used in a wide range of applications, ranging from medical to personal, such as Parkinson’s
disease monitoring, rehabilitation, microsurgical devices, fall detection and fitness . Recently,
the need for increased recognition capabilities paved the way to the exploitation of AI
paradigms. They are starting to be deployed in HAR by using Pattern Recognition (PR) models
(decision tree, support vector machine, naïve Bayes) or Deep Learning (DL) models . Although
DL enables superior recognition capabilities and accuracy than PR , the latter is often
employed in resource constrained devices for its lower computational complexity . Cloud
computing could be used to move away part of the computational load of DL from local
devices . However, it is not an optimal solution because of the additional energy budget
required for communications, and the performance degradation due to bandwidth limitations.
Therefore, keeping the computation close to the sensing element, according to the edge
computing paradigm, appears the most viable solution to minimize delay and power
consumption .
3. Problem statement
Humans impact the physical environment in many
ways: overpopulation, pollution, burning fossil fuels, and
deforestation. Changes like these have triggered climate
change, soil erosion, poor air quality, and undrinkable water.
4. Motivation
Understanding people actions and their interactions with the environment is a key
element for the development of the aforementioned intelligent system. Human activity
Recognition is a field that specifically deals with this issues through the integration of
sensing and reasoning, in order to deliver context aware data that can be employed to
provide personalized support in many applications.
It is possible to infer the activities performed by its residents based on sensors signals
along with other relevant aspects such as time of the day and date .
5. Objective
The goal of human activity recognition is to examine
activities from video sequences or still images. Motivated
by this fact, human activity recognition systems aim to
correctly classify input data into its underlying activity
category.
6. Literature Survey
Sr.no Title Author Abstract
1 Multi-view
human activity
recognition
using motion
frequency
Neslihan
Käse; Mohammadre
za Babaee; Gerhard
Rigoll
The problem of human activity recognition
can be approached using spatio-temporal
variations in successive video frames. In this
paper, a new human activity recognition
technique is proposed using multi-view
videos. Initially, a naive background
subtraction using frame differencing between
adjacent frames of a video is performed.
Then, the motion information of each pixel is
recorded in binary indicating existence/non-
existence of motion in the frame. A pixel
wise sum over all the difference images in a
view gives the frequency of motion in each
pixel throughout the clip.
7. Literature Survey
Sr.no Title Author Abstract
2 Demo: Hands-
Free Human
Activity
Recognition
Using
Millimeter-
Wave Sensors
Soo Min
Kwon; Song
Yang; Jian Liu; Xin
Yang; Wesam
Saleh; Shreya
Patel; Christine
Mathews; Yingying
Chen
In this demo, we introduce a hands-free
human activity recognition framework
leveraging millimeter-wave (mmWave)
sensors. Compared to other existing
approaches, our network protects user
privacy and can remodel a human skeleton
performing the activity. Moreover, we show
that our network can be achieved in one
architecture, and be further optimized to
have higher accuracy than those that can
only get singular results (i.e. only get pose
estimation or activity recognition). To
demonstrate the practicality and robustness
of our model, we will demonstrate our model
in different settings (i.e. facing different
backgrounds) and effectively show the
accuracy of our network.
8. Literature Survey
Sr.no Title Author Abstract
3 A Hybrid
Approach for
Human
Activity
Recognition
with Support
Vector
Machine and
1D
Convolutional
Neural
Network
Md Maruf Hossain
Shuvo; Nafis
Ahmed; Koundinya
Nouduri; Kannapp
an Palaniappan
The Human Activity Recognition (HAR) is
a pattern recognition task that learns to
identify human physical activities
recorded by different sensor modalities.
The application areas include human
behavior analysis, ambient assistive living,
surveillance-based security, gesture
recognition, and context-aware
computing. The HAR remains challenging
as the sensor data is noisy in nature and
the activity signal varies from person to
person. To recognize different types of
activity with a single classifier is often
error-prone.
9. Literature Survey
Sr.no Title Author Abstract
4 Automated
daily human
activity
recognition for
video
surveillance
using neural
network
Mohanad
Babiker; Othman
O. Khalifa; Kyaw
Kyaw Htike; Aisha
Hassan; Muhamed
Zaharadeen
Surveillance video systems are gaining
increasing attention in the field of
computer vision due to its demands of
users for the seek of security. It is
promising to observe the human
movement and predict such kind of sense
of movements. The need arises to develop
a surveillance system that capable to
overcome the shortcoming of depending
on the human resource to stay
monitoring, observing the normal and
suspect event all the time without any
absent mind and to facilitate the control
of huge surveillance system network.
10. Literature Survey
Sr.no Title Author Abstract
5 Optimal Time-
Window
Derivation for
Human-Activity
Recognition
Based on
Convolutional
Neural
Networks of
Repeated
Rehabilitation
Motions
Kyoung-Soub
Lee; Sanghoon
Chae; Hyung-Soon
Park
This paper analyses the time-window size
required to achieve the highest accuracy of
the convolutional neural network (CNN) in
classifying periodic upper limb
rehabilitation. To classify real-time motions
by using CNN-based human activity
recognition (HAR), data must be segmented
using a time window. In particular, for the
repetitive rehabilitation tasks, the
relationship between the period of the
repetitive tasks and optimal size of the time
window must be analyzed. In this study, we
constructed a data-collection system
composed of a smartwatch and smartphone.
Five upper limb rehabilitation motions were
measured for various periods to classify the
rehabilitation motions for a particular time-
window size.
12. APPLICATIONS
1) Applicable for the end users such as fall detection, behavior-based context-
awareness , home and work automation, and self-managing system;
2) Applications for the third parties such as targeted advertising, research platforms for
the data collection, corporate management and accounting;
3) Applications for the crowds and groups such as social networking and activity-based
crowdsourcing.
13. ADVANTAGES
Analyze the activity of a person from the information collected .
Discover activity pattern that determine which activity is doing a person.
Calculate a predictive model that can recognize a person's activity.
Design individual exercise table to improve the health of person.
14. EXESTING SYSTEM DRAWBACKS
In the proposed approach, the difference images between adjacent frames
of a video are used to find motion barcodes as in [2] for each pixel. Then for
each view, a motion frequency of each pixel is determined by cumulative
sum over all difference images of one view. For each camera view, we obtain
one pixel motion frequency vector. Motion frequency vector of all camera
views are concatenated together to get discriminative features. Pixel motion
frequency vector provides information about frequency of presence/absence
of motion in a particular pixel.
15. CONCLUSION
In this article, a new HNN model and a custom HW accelerator for HAR applications
has been proposed. The system obtains high accuracy with low power consumption and
resources compared to the state-of-the-art. Results show that the system is able to
accomplish its task with a minimum power consumption of 6.3 μW and an area
occupation of 0.2 mm2, that is promising for the integration of the accelerator in the
same die with the sensing element, thus realizing an AI-based edge device. Future
works will be aimed to the extension of the architecture to other applications requiring
high operation frequencies and the possibility to use ternary weights.
16. References
1. Haleem A., Javaid M., Khan I.H. Internet of things (IoT) applications in
orthopaedics. J. Clin. Orthop . Trauma. 2019 doi: 10.1016/j.jcot.2019.07.003. In press. [PMC
free article] [PubMed] [CrossRef] [Google Scholar]
2. Bai L., Yang D., Wang X., Tong L., Zhu X., Bai C. Chinese experts’ consensus on the
Internet of Things-aided diagnosis and treatment of coronavirus disease 2019. Clinical
eHealth. 2020 Mar 17 In press. [Google Scholar]
3. Wang Y, Hu M, Li Q, Zhang XP, Zhai G, Yao N. Abnormal respiratory patterns classifier
may contribute to large-scale screening of people infected with COVID-19 in an accurate and
unobtrusive manner. arXiv preprint arXiv:2002.05534. 2020 Feb 12.
4. Haleem A., Javaid M., Vaishya R., Deshmukh S.G. Areas of academic research with the
impact of COVID-19. AJEM (Am J Emerg Med) 2020 doi: 10.1016/j.ajem.2020.04.022. In
press. [PMC free article] [PubMed] [CrossRef] [Google Scholar]
5. Mohammed M.N., Syamsudin H., Al-Zubaidi S., A K S, Ramli R., Yusuf E. Novel COVID-
19 detection and diagnosis system using IOT based smart helmet. Int J Psychosoc
Rehabil. 2020;24(7) [Google Scholar]
