GESTURE VOLUME CONTROL USING OTSU METHOD

Abdul Subhani Shaik1, Suresh Merugu1, Sorup Dutta1 and Surendra Pal Singh2
1
Department of Electronics and Communication Engineering,
CMR College of Engineering & Technology, Medchal Road, Hyderabad, Telangana, India.
2
Faculty of Engineering and Technology, Wollega University, Nekemte City,
PO Box- 395, Ethiopia.
[email protected] and [email protected]

ABSTRACT:

In current world of automated systems, hand gesture technology is used not only in gaming but also in medical
applications, industrial applications, IT hubs, banking sectors, and so on. This project is associated with a
particular concept of a laptop or computer with hand gesture control. The Human Machine Interface (HMI) is a
system comprised of hardware and software that aids in the establishment of communication and information
exchange seen between user and the machine. A hand gesture recognition system enables nonverbal communication
that is natural, innovative, and modern. This implementation's goal is to discuss a novel approach to hand gesture
recognition method for detection of such shape-based features. The layout shows a single camera that captures the
user's gesture and feeds it into the system. The main objective of gesture recognition is to develop a system that can
recognize specific human gestures and use them to communicate information for system control. By moral
responsibilities gesture recognition, a person can access a computer by performing a particular gesture in front of a
video camera connected to a computer. In this project, we will use the OpenCV module to create a hand gesture
volume control system. The system can be operated using hand gestures rather than a keyboard and mouse in this
case.

KEYWORDS: Hand Gesture, Detection, OpenCV, Recognition, Control, HMI, Otsu Method, Computer Vision

1. INTRODUCTION:

Image processing, which interacts with extraction of useful information from a digital image, has played a
distinctive role in the advancement of technological advancements in recent years. It is divided into two tasks:
improving pictorial information for human interpretation and processing image data for storage, transmission, and
recognition [1] for independent computer vision. Today, almost no field of technical Endeavour is unaffected by
digital image processing. Countless application-oriented image analyzers are accessible and work well in the real
world. Hand gestures are a powerful and spontaneous mode of communication for Human Computer Interaction
(HCI) [6].

Gestures are expressive, meaningful bodily motions that involve physical movements of the fingers, hands,
arms, head, face, or torso with the goal of communicating meaningful information or interacting with the
environment. They make up a fascinating tiny subspace of conceivable human movement. The environment may
interpret a gesture as a compression mechanism for information to be conveyed elsewhere and then reassembled by
the recipient. Many-to-one mappings from concepts to gestures and vice versa are common. As a result, gestures are
imprecise and undefined. To convey the concept "halt," for example, a raised hand with palm facing forward or an
excessive wave of both hands over the head can be used. Gestures, like speech and handwriting, differ from one
person to the next, and even from one occurrence to the next. Static gestures and dynamic gestures [2] (including
prestrike, stroke, and poststroke phases) are both possible. As with sign languages, certain gestures contain both
static and dynamic features. Automatic recognition of natural continuous gestures, yet again, necessitates temporal
segmentation. Often, the start and end points of a gesture must be specified in terms of movement frames, both in
time and space [3]. The background of previous and following gestures can sometimes influence a gesture.
Furthermore, gestures are often culturally and linguistically distinctive. They can fall under one of the following
categories:
1) Hand gestures: sign languages, and entertainment applications

2) Face gestures: a) head movements; b) vision-based direction; c) lifting the eyebrows; d) jaw movement; e)
blinking, f) nose movement and g) facial expression

3) Body gestures: full-body motion is involved, as in: a) tracking the movements of two individuals engaging
outside; b) evaluating a dancer's movements to generate matching music and graphics; and c) detecting human
postures for patients recovering and athletic training.

Extraction of sensitive features from hand landmarks such as regions surrounding the finger tips, knuckle
movement of a normalized image [5] is required. To develop relevant characteristics, dynamic picture frames of
these locations are frequently tracked. For recognizing an expression, the location, intensity, and dynamics of
hand gestures are crucial. Furthermore, measuring the strength of spontaneous hand gestures is generally more
challenging than measuring the intensity of hand position [21─23]. Hand tension, general muscle tension, self-
contact areas are some of the more subtle indicators that are sometimes used.

Hand tracking and segmentation are the first steps in any hand gesture recognition process. The proposed system
will consist of desktop/laptop interface which use web camera for capturing hand gesture. Using this communication
mode, we can control the volume of a system. This system is intended to supplement biological vision by describing
artificial vision systems implemented in software [4].

2. LITERATURE SURVEY:

Karishma et al. (2014) proposed Hand Gesture Segmentation using a Combination of Skin Color Detection and
Background Subtraction recommended an algorithm that combines a skin color model and background subtraction
to produce robust outcome in the availability of extreme lighting changes. If there is no important grayscale
variation or a coincide of grayscale pixel values, obtaining accurate segments becomes extremely difficult [9].

Dhawan et al. (2013) In Hand Detection-Based Techniques for Human-Computer Interaction, hand detection and
segmentation is used. The main drawback is it allows regions to overlap. To overcome this drawback contour
method can be used [10].

S. Gokturk et al. (2001) Application of Region of Interest Based Medical Image Compression to Colon CT Images
discusses a lossless hybrid model in the area of interest, with greater, motion-compensated lossy compression in the
other areas but the quality of the image is reduced. It should work more on the image enhancement [11].

Soo et al. (2014) proposed Detecting objects with the Haar-cascade Classifier is a computer learning-based approach
which uses a large number of positive and negative images to perform the classification. The drawback in it is
accuracy is less and unfit bounding boxes. If the accuracy is improved, we could get the perfect output [12].

Nahar and Ali et al. (2014) proposed an Improved Approach for Digital Image Edge Detection uses discontinuous
local image features to identify edges and thus define an object's boundary because there are too many edges in the
picture and there isn't enough contrast between objects, this technique isn't appropriate. Hence need to apply another
technique which would be suitable [14].

Nayana et al. (2014) proposed solution is Hand Gesture Recognition Technique Implication for HCI Using Open CV
which tells to separate the hand from the background image, a histogram-based approach is used. To achieve the
best results, background termination techniques are used. The limitation in it is it doesn’t allow Not allow you to
read exact values because data is grouped into categories [15].

Direkoglu et al. (2008) proposed solution is Gaussian filter-based image-based multi - resolution shape description
in which it creates a new multi resolution Fourier-based image description in two-dimensional space by using a low-
pass Gaussian function and a high-pass Gaussian filter individually. The Proposed method is non-causal, which
means that the filter screen in the time domain is geometric about the origin. As a result, the Gaussian filter is
physically impossible to implement [16].

Velastin et al. (2001) For underground platforms, an automatic congestion detection system is used. When
overpopulation is identified, the system notifies station technicians, who must take immediate measures to avoid
accidents such as falling down or being tried to push onto the routes. The major disadvantage is the inaccuracy, as
the gradient magnitude of the edges decreases [17].

Viola et al. (2001) proposed a boosted cascade of simple features is used to detect objects quickly which is a
machine learning approach for visual object detection capable of processing images at high detection rates and
processing images at high speeds. Drawback about this is it cannot detect people as accuracy is low [13]. Figure 1
shows the block diagram of the proposed method.

3. METHODOLOGY:

Image Acquisition Pre- Hand Region

processing Segmentation

Controlled Volume Classified Hand Posture

Output Gesture Recognition

Fig.1. Block diagram of the proposed method

3.1. Image acquisition and preprocessing

We take camera as an input for image acquisition. For preprocessing unwanted noises are removed for the
improvement of image and can enhance it.

a. Gray scale conversion: In this case, a 3-channel image (RGB) is used, which means that we must add r, g, and b
and then divide by 3 to obtain the desired grayscale image.

b. Noise Reduction: To remove/reduce the noise, we have used Gaussian blur filter which helps in image size
reduction and for better edge detection. It also smoothens the image.

Figure 2 shows the RGB Image, Gray scale image and Gaussian Blur Image.
 (a) (b) (c)
Fig.2. (a) RGB image (b) Gray Scale Image (c) Gaussian Blur image

3.2. Image Segmentation (Threshold Segmentation)

It's being used to perform image thresholding automatically. The algorithm, in its most basic form, returns a single
intensity threshold that divides pixels into two classes: foreground and background. The separation process
foreground pixels from background pixels are known as thresholding. There are numerous methods for achieving
optimal thresholding, one of which is known as the Otsu's method. Otsu's method is a significant variation technique
for determining the threshold value with the least weighted variance between the foreground and background pixels.
The main idea here is to repeat through all feasible threshold values and measure the expansion of background and
foreground pixels. Then, locate the threshold with the smallest spread [18][24].

Given an 8-bit grayscale image I and the image pixel Pi of the i-th pixel, the number of repetitions of pixel value [0,
255] can be used to generate a grayscale histogram H. Otsu thresholding uses this statistical data directly to
determine the optimal cut that provides the best detachment of the classes (for image binarization, foreground, and
background). Figure 3 shows the Otsu Thresholding method.

Fig.3. Otsu Thresholding

First, create the histogram H algorithmically (colored in blue). Furthermore, as shown in fig 3 for each threshold t in
[0, 255], pixels can be divided into two classes, C1 and C2; pixels with Pi t are placed in C1, while those with Pi t
are placed in C2. The prospects of C1 and C2 separated by t are then denoted as W1 and W2, respectively. For
example, W1 = (number of pixels in C1) / (total pixels count). Third, given H, W1, and W2, compute the between-
class variance Vb or within-class variance Vw for each t. Finally, the optimal cut t* relates to t for which Vb or Vw
is maximum. According to the original paper [19] the author introduces "class separability" (CS) to analyze the
goodness of the thresholding result. For Maximum Vb; C1, C2, and C3 contain black (grayscale value = 0), grey
(grayscale value = 128), and white (grayscale value = 255) colors, respectively.

Specifically, because of the difference in grayscale values between the two classes (at this point, you can think of it
as Vb). Assume one class is the foreground and the other is the background. C1 and C3, because they have the most
visual distinction between the two classes. As a result, higher Vb is preferred for binarization. For minimum Vw;
Given two separation cases, C1 is pure black and C2 is pure white, and C1 and C2 with grey added. The former,
because each element is more similar to other elements in the same class than those in the latter. In other words, the
total variance of elements in C1 and C2 (denoted as V1 and V2) in the former case is less than that in the latter (at
this point, V1+V2 can be thought of as Vw). This is the reason why a lower Vb is a better choice for binarization.
Let’s talk more about the class separability (CS, for short) and Vb (and Vw).

(a) (b)
Fig.4. (a) Good separability (b) Not so good separability

As in the figure 4, we can say that left side has good separability because

(1) the classes (blue and yellow points) are farther apart, and
(2) each element in each class is most focused.
Strangely, (1) indicates maximum Vb and (2) indicates minimum Vw. Fisher’s linear discriminant analysis [8] gave
rise to the term CS. Otsu thresholding is equivalent to Vb/Vw and can be thought of as a discrete version of CS.
Remember that a higher CS indicates better thresholding results. As a result, the larger Vb and smaller Vw,
consistent with reasons (1) and (2), can increase the value CS. Because the variance is either between or within the
class, the total variance Vt is held constant (i.e., Vt = Vb + Vw). Vb/Vw can be rewritten as Vb/(Vt-Vb) or (Vt-
Vw)/Vw. In that equation, the larger Vb contributes the same amount to the smaller Vw, so either of the two reasons
(1) or (2) can result in a higher CS. So far, we understand why the best cut for image binarization in Otsu
thresholding is on the threshold t with the highest Vb (Vw is minimum). To the best of my knowledge, there is little
online material discussing how to derive Vb and Vw. Returning to the image domain now, may have come across
the equation Vw = W1* V1 + W2 *V2.
 V w =W 1∗V 1 +W 2∗V 2 (1)

∑ π2 ∑ π2
πϵ C 1 πϵ C 2 (2)
−W 1∗μ21 + −W 2∗μ 22
N N

In this section, I define each term, where N is the total pixel count in I and W1 and W2 are the probabilities of the
two classes C1 and C2, separated by t. C1 contains pixels with values less than t, whereas C2 contains pixels that do
not belong in C1. Because C1’s size may differ from C2, we must convert V1+V2 to a weighted variance sum for
Vw.

2
V b =W 1∗W 2∗(μ 1−μ2) (3)

Vb is proportional to the difference between class representatives in the second equation (as mentioned in (1)
Maximum Vb). Unlike Vw, it is difficult to define Vb mathematically. However, we can do back derivation using
Vt and Vw.

(∑ )
N
π2
i=0 (4)
V t= −μ t2
N

First, expand the squared total mean, which is highlighted in blue, Second, for t, substitute the weighted sum of 1
and 2. Third, to merge and arrange the expansion terms, use W1+W2=1. We can derive the two equations from
these.

2 2 2 2
V b =W 1∗μ1 +W 2∗μ2−2∗μt + μt (5)

The second derived equation yields two observations:

1) Given constant W1 x W2, a large mean difference encourages greater Vb.

2) Maximum Vb occurs when W1 = W2 for a fixed mean difference.
In the above observations, Vb forms a push-pull factor during maximization, with a push between 1 and 2 and a pull
between W1 and W2. Furthermore, during threshold finding in Otsu binarization, a pull exists within each class to
concentrate elements and a push exists between the two classes to allow one to be separated from the other.

In the proposed methodology, we have used Otsu method in which the following steps are involved:
 At first, we process the input image which is a RGB image.
 Then we obtain the image histogram which is distribution of pixels.
 Next, we divide the distribution of pixels into classes.
 We calculate the intra variance among the pixels of same class and inter variance between two classes.
 Repeat the above step until the intra-variance is minimum or inter-variance is maximum
 Calculate the threshold value which is the sum of class variances with their associated weights
 Replace image pixels into white in those regions, where pixel intensity is greater than threshold and into the
black in the opposite cases.
 Fig.5. Image Thresholding

Figure 5 shows the Image Thresholding.

3.3. Hand detection and tracking:

 We're using OpenCV to identify human hands and calculate the distance between both the palm centre and
the finger tips.
 We're using the contour tool, which is similar to finding a white object against a black background. As a
result, the object to be observed should be white, and the background must be black. Fig.6. shows otsu
segmentation, binary image, contour detection with defects, contour detection without defects.

(a) (b) (c) (d)

Fig.6. (a) Otsu Segmentation (b) Binary Image (c) Contour detection with defects (d) Contour detection without
defects

3.4. Hand Posture Recognition:

We are using Support Vector Machine (SVM) model that uses classification algorithms for two group classification
problems. The SVM algorithm's goal is to find best line or classification algorithm that can split n-dimensional
space into categories so that new data points can be easily placed in the relevant category in the upcoming years.
The best decision boundary is a hyperplane. Fig.7. shows hand detection and tracking, hand posture recognition
using SVM, Featuring advantages of proposed method.
 (a) (b) (c)
Fig.7. (a) Hand detection & tracking (b) Hand posture recognition using SVM (c) Featuring advantages of proposed
method.

Advantages:

 It will be able to detect one hand at once which means if both the hands are shown in front of camera, it
will take any one which will be fixed and will have access.
 High portability and reduction of external devices like keyboard and mouse.

Table-1 shows the comparison of the state-of -the-art works of Blanca Miriam Lee-Cosio et al. [25], MM Gharasuie
et al. [26], Zhou Ren et al. [27], Joseph C. Tsai et al. [20] with our proposed model, which produced better results
with higher accuracy.

Table-1: Comparison of various works with our proposed model

Reference Model Dataset Accuracy

Blanca Miriam Lee-Cosio Fast Fourier Transform, 15 clusters 95%

et. al. [25] K-means algorithm

MM Gharasuie Baum-Welch Algorithm 210 instances of 7 93%

et al. [26] hand gesture set
Zhou Ren Finger Earth Movers 10 gesture datasets 94%
et al. [27] Distance Algorithm
Joseph C. Tsai Camshift Algorithm 700 image set 91%
et al. [20]
Proposed Model Otsu Model 150 hand gesture 96%
set

3.5. Hardware and Software Requirements:

The main hardware requirement is Personal computer (PC) or laptop which has at least Intel I3, Dual core
processor, 4GB RAM. Need to have free disk space as it will be affected by the size of the partition and the
installation of online help files.
OpenCV Version-4.4.0.44 which has at least I3 processor, 4GB RAM and has 2GB disk space available. Mediapipe
Version-0.8.8.1 which allows to run variety of machine learning models in real time [7][21] which helps to get
accurate readings. Python Version-3.7 is used as this version supports OpenCV and Mediapipe as of now. Later new
versions would be helpful to get more outcomes. Core Audio Window Library is used to play audio on PC or laptop.
Usually, we have built-in software for it but in some systems it doesn’t have. So need to add this if it is not
available.

4. RESULTS AND DISCUSSIONS:

Experimental Results:

(a) (b)
Fig.8. (a) Volume unchanged due to unrecognized posture (b) Full Volume Recognized Posture

For tracking, the suggested model will use a hand bounding box around the palm. To give extra oversight on the
basis of hand topology and to better cover various hand positions, 21 3D coordinates are annotated on palm. In
Fig.8. (a), the model can identify and track the hand but it cannot be utilized to regulate the system's volume
because the little finger is up. In this model, we use to little finger as a key to alter the volume. In Fig.8. (b), the
volume level is proportional to the distance between the fore and thumb fingers when the little finger is lowered.
The distance bar delineates the percentage stretch between the fore and thumb fingers, as well as the volume level
in an oblique way. Table 2 shows the Comparison of different parameters of surveyed Hand Gesture Recognition
systems

Table 2: Comparison of different parameters of surveyed Hand Gesture Recognition systems

Method Background Segmentation Feature Vector Classifier/ Accuracy

technique Representation Recognition
[Ref. (%)
No.]
Euclidean
[28] Uniform HSV colour distance between SVM
space centroid of hand NA
and fingertip
[29] Uniform Pixel intensity Haar-like features Principal Hand –
 Component 94.63 Face
thresholding
Analysis (PCA) – 98.4
Mean Euclidean
[30] Single 92.7
Uniform Thresholding distance, mean
Probabilistic model Double 96.2
turning angle

[31] Manually from Blob and ridge

Cluttered Particle filtering 86.5
the background features

Color-based Bi-driven inference

[32] Uniform region of- Active contour (Dynamic Bayesian 95
interest Network + HMM)
Bayesian rule Euclidean
[33] Mostly uniform based RGB distance and Distance classifier 93.38
color angles
Euclidean
[34] Color based distance from the
Uniform
HSV boundary, centre
NA NA
of palm
[35] Depth
Mostly uniform Sparse depth map HMM 94
thresholding
Iterative Closest
Point (ICP)
[36] RGB of skin
Cluttered Angular pose matching, Voxel
color
quantization and
NA
binary matching
YCrCb color Location,
[37] Fully cluttered space and 3D orientation and HMM 92.3
depth map velocity
[38] YCrCb color
Mostly uniform HU moment Contour matching 89.53
space
Location,
Otsu orientation and
[39] Cluttered Thresholding of velocity of HMM 85
skin - color centroid of the
hand

5. CONCLUSION:
The proposed work will contribute to the complete elimination of the hardware components. It takes a webcam to
capture an Input image and we perform certain methodology to get desired output. This would create a new
generation of human-computer interaction in which there would be no need for physical contact with the device.
The system allows anyone to easily operate a computer through using gesture instructions. The experimentation
is carried out on a variety of datasets, demonstrating that the suggested technique outperforms previous methods
by being resistant to scale variance and requiring no specified templates for recognition.

6. REFERENCES:

[1] Akira Utsumi, Tsutoniu Miyasato,Fumio Kishinoand Ryohei Nakatsu, “Real-time Hand Gesture Recognition System,”
Proc. of ACCV ’95, vol. 11, pp. 249-253,Singapore,1995
[2] Attila Licsár, TamásSzirányi University of Veszprém, “Dynamic Training of Hand Gesture Recognition System”
Department of Image Processing and Neurocomputing, H-8200 Veszpré, 23- 26 Aug. 2004Harguess, J., Aggarwal, J.K.,
A case for the average-halfface in 2D and 3D for face recognition, IEEE Computer Society Conference on Computer
Vision and Pattern Recognition Workshops, pp. 7-12, 2009.
[3] L. Bretzner and T. Lindeberg, “Relative orientation from extended sequences of sparse point and line correspondences
using the affine trifocal tensor,” in Proc. 5th Eur. Conf. Computer Vision, Berlin, Germany, June 1998, vol. 1406,
Lecture Notes in Computer Science, pp.141–157, Springer Verlag
[4] Suresh Merugu, Anuj Tiwari & Surendra Kumar Sharma, (2021), Spatial–Spectral Image Classification with Edge
Preserving Method, Journal of the Indian Society of Remote Sensing ISSN 0255-660X J Indian Soc Remote Sens DOI
10.1007/s12524-020-01265-7.
[5] G. R. S. Murthy, R. S. Jadon. (2009). “A Review of Vision Based Hand Gestures Recognition,” International Journal of
Information Technology and Knowledge Management, vol. 2(2), pp. 405410.
[6] Mokhtar M. Hasan, Pramoud K. Misra, (2011). “Brightness Factor Matching For Gesture Recognition System Using
Scaled Normalization”, International Journal of Computer Science & Information Technology (IJCSIT), Vol. 3(2).
[7] Shaik, A.S., Karsh, R.K., Islam, M. et al. A review of hashing based image authentication techniques. Multimed Tools
Appl 81, 2489–2516 (2022). https://doi.org/10.1007/s11042-021-11649-7
[8] S. Y. Chen, Y. F. Li, and J. W. Zhang, “Vision processing for real-time 3D data acquisition based on coded structured
light,” IEEE Trans. Image Process., vol. 17, no. 2, pp. 167–176, FebMartinez,
[9] Shaik, A.S., Ahmed, M., Suresh, M. (2022). Low-Cost Irrigation and Laser Fencing Surveillance System for the Paddy
Fields. Advances in Micro-Electronics, Embedded Systems and IoT. Lecture Notes in Electrical Engineering, vol 838.
Springer, Singapore.
[10] S. N. Karishma and V. Lathasree, “Fusion of Skin Color Detection and Background Subtraction for Hand Gesture
Segmentation”, (International Journal of Engineering Research and Technology) vol. 3 no 1 pp 13–18, 2014.
[11] Shaik, A.S., Karsh, R.K., Suresh, M., Gunjan, V.K. (2022). LWT-DCT Based Image Hashing for Tampering
Localization via Blind Geometric Correction. ICDSMLA 2020. Lecture Notes in Electrical Engineering, vol 783.
Springer, Singapore. https://doi.org/10.1007/978-981-16-3690-5_156
[12] S. Gokturk C ,Tomasi B and Girod C. Beaulieu, “Region of Interest Based Medical Image Compression with
Application to Colon Ct Images”, (International Conference of the IEEE Engineering in Medicine and Biology Society)
vol 23 pp 575-578, 2001
[13] S. Soo, “Object detection using Haar-cascade Classifier”, (Inst. Comput. Sci. Univ. Tartu) vol. 2 no. 3 pp 1–12, 2014
[14] P. Viola and M. Jones, “Rapid object detection using a boosted cascade of simple features”, (IEEE Comput. Soc. Conf.
Comput. Vis. Pattern Recognition. CVPR 2001) vol. 1 pp 1511- 1518, 2001
[15] S Merugu, K Jain, A Mittal, B Raman, (2020), Sub-scene Target Detection and Recognition Using Deep Learning
Convolution Neural Networks - ICDSMLA 2020, Lecture Notes in Electrical Engineering, Springer, Singapore, pp
1082-1101.
[16] Devadasu, G., Sushama, M., Identification of voltage quality problems under different types of Sag/Swell faults with
Fast Fourier Transform analysis, Proceeding of IEEE - 2nd International Conference on Advances in Electrical,
Electronics, Information, Communication and Bio-Informatics, IEEE – AEEICB 2016, pp. 464–469.
[17] Merugu S., Reddy M.C.S., Goyal E., Piplani L. (2019), Text Message Classification Using Supervised Machine
Learning Algorithms. In: Kumar A., Mozar S. (eds) ICCCE 2018. Lecture Notes in Electrical Engineering, ISSN 1876-
1100, Vol. 500. Springer, Singapore.
[18] P. B. Nayana and S. Kubakaddi, “Implentation of Hand Gesture Recognition Technique for HCI Using Open CV”,
(International Journal of Recent Dev) vol. 2 no. 5 pp 17–21, 2014
[19] Direkoglu, C. and M.S. Nixon, “Image-based multiscale shape description using Gaussian filter”, IEEE Xplore Press,
Bhubaneswar, pp: 673-678, 2008
[20] B.P.L. and S.A. Velastin, “Automatic congestion detection system for underground platforms”, IEEE Xplore Press,
Hong Kong, pp: 158-161, 2001
[21] Otsu, Nobuyuki. “A threshold selection method from gray-level histograms.” IEEE transactions on systems, man, and
cybernetics 9.1 (1979): 62–66.
[22] V. I. Pavlovic, R. Sharma, and T. S. Huang, “Visual interpretation of hand gestures for human computer interaction,”
IEEE Trans. Pattern Anal. Mach. Intell., vol. 19, no. 7, pp. 677–695, Jul. 1997.
[23] Joseph C. Tsai, Shih-Ming Chang, Shwu-Huey Yen, Kuan-Ching Li, Yung-Hui Chen,Timothy K. Shih. “A real-time
hand gesture recognition system for daily information retrieval from the internet”, International Journal of
Computational Science and Engineering Vol. 11, No. 2,2015
[24] Premalatha, B., Srikanth, G., Babu, P.R., Sonth, M.V.” Design and analysis of two dimensional electromagnetic band
 gap antenna for WIFI applications”, AIP Conference Proceedings, 2021, 2358, 090005.
[25] Syed, S Merugu, V Kumar, (2020), Augmented Reality on Sudoku Puzzle Using Computer Vision and Deep Learning, -
Advances in Cybernetics, Cognition, and Machine, 2020, Lecture Notes in Electrical Engineering, Springer, Singapore,
pp 567-578.
[26] H. S. Yoon, J. Soh, Y. J. Bae, and H. S. Yang, “Hand gesture recognition using combined features of location, angle and
velocity,” Pattern Recogn., vol. 34, pp. 1491–1501, 2001
[27] Shaik, A.S., Karsh, R.K., Islam, M. (2021). Robust Image Hashing Using Chromatic Channel. Proceeding of Fifth
International Conference on Microelectronics, Computing and Communication Systems. Lecture Notes in Electrical
Engineering, vol 748. Springer, Singapore.
[28] R. Bowden, A. Zisserman, T. Kadir, and M. Brady, “Vision based interpretation of natural sign languages,” in Proc. 3rd
Int. Conf. Comput. Vis. Syst., Graz, Austria, Apr. 2003, pp. 391–401.
[29] Devadasu, G., Sushama, M., A novel multiple fault identification with fast fourier transform analysis 1st International
Conference on Emerging Trends in Engineering, Technology and Science, ICETETS, 2016.
[30] J. Daugman, “Face and gesture recognition: An overview,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 19, no. 7, pp.
675–676, Jul. 1997.
[31] Merugu S., Reddy M.C.S., Goyal E., Piplani L. (2019), Text Message Classification Using Supervised Machine
Learning Algorithms. In: Kumar A., Mozar S. (eds) ICCCE 2018. Lecture Notes in Electrical Engineering, ISSN 1876-
1100, Vol. 500. Springer, Singapore.
[32] Blanca Miriam Lee-Cosio, Carlos Delgado-Mata, Jesus Ibanez,“ANN for gesture recognition using accelerometer data”,
Procedia Technology 3, 109-120, 2012
[33] MM Gharasuie, H Seyedarabi,“Real-time dynamic hand gesture recognition using hidden Markov models”, 8th Iranian
Conference on Machine Vision and Image Processing (MVIP), 194-199, 2013
[34] Zhou Ren, Junsong Yuan, Jingjing Meng, Zhengyou Zhang , “Robust part-based hand gesture recognition using kinect
sensor”,IEEE transactions on multimedia 15 (5), 1110-1120, 2013
[35] Pradhan, A., Ghose, M. K. and Pradhan, M. Hand Gesture Recognition using Feature Extraction. Int. J. of Current Engg.
And Tech. 2(4) (Dec. 2012), 323 – 327
[36] Bui, T. T. T., Phan, N. H. and Spitsyn, V. G. 2012. Face and Hand Gesture Recognition Algorithm Based on Wavelet
transforms and Principal Component Analysis. In Proc. of 7th Int. Conf. on Strategic Technology.
[37] Merugu, S., & Ghinea, G. (2022). A Review of Some Assistive Tools and their Limitations for Visually Impaired. Helix
- The Scientific Explorer | Peer Reviewed Bimonthly International Journal, 12(1), 1-9.
[38] Premalatha, B., Prasad, M.V.S. & Murthy, M.B.R. Multi-Band Notched Antennas for UWB Applications.
Radioelectron.Commun.Syst. 62, 609–618 (2019). https://doi.org/10.3103/S073527271912001X.
[39] Anand, R., Muthubalaji, S., Karuppiah, N., Devadasu, G., Performance enhancement analysis of 7-level symmetric
multilevel inverter with multi-carrier pulse width modulation, Journal of Advanced Research in Dynamical and Control
Systemsthis, 2018, 10(14 Special Issue), pp. 856–863