Emotion Detection Algorithm Report

This report documents the development of an emotion detection algorithm for a music player. The system aims to automate music selection based on the user's emotional state, detected through facial expressions using a webcam. The report details the progress made using three core algorithms: Support Vector Machines (SVM), edge detection (improved Canny operator), and face detection (improved Viola-Jones). Each algorithm's performance, challenges encountered, and planned future work are discussed. The report includes a literature review citing relevant research on emotion recognition, image processing, and machine learning techniques. Appendices provide visual aids illustrating hyper-plane vectors and support vectors. The conclusion summarizes the project's progress and suggests future improvements for the emotion-based music player.

Emotion Detection Algorithm Player

Table of Contents
Introduction and overview...........................................................................................................................2
Progress to date............................................................................................................................................3
SVM:........................................................................................................................................................3
Performance:............................................................................................................................................3
Edge detection:........................................................................................................................................4
Problems:.................................................................................................................................................4
Performance:............................................................................................................................................5
Face detection:.........................................................................................................................................5
Problems:.................................................................................................................................................5
Performance:............................................................................................................................................6
Planned work...............................................................................................................................................6
Steps in SVM:..........................................................................................................................................6
Steps in edge detection:...........................................................................................................................7
Steps in face detection:............................................................................................................................7
Conclusion and Recommendations..............................................................................................................8
Bibliography................................................................................................................................................9
Appendices:...............................................................................................................................................11
1

Introduction and overview
Music is an important part of the entertainment. Manual work is optimized and it gains huge attention
for technological advancement. At present, there are various traditional music players, which select and
organize the songs manually. User needs to create and modify the play-list according to his mood. It is a
time-consuming process. Some music players have included advanced features such as lyrics and
recommendation according to the preferred genre or singer. The user enjoys some of the features, and
there is room for improving the automation field in case of music players. Selection of songs
automatically and the arrangement depends on the user’s mood for a better experience (Kabani et al.,
2015). This is accomplished in the system that reacts to the user’s emotions, saves time that is spent in
entering data manually.
Emotions are expressed by facial expressions, gestures, speech etc. The system understands the user’s
mood from his facial expression. The system is capable of capturing facial expression by utilizing the
device’s camera. There are several emotion recognition systems, which takes image, speech etc. as input
for determining the emotion. The different approaches used in image processing edge detection are the
emotional mouse, magic pointing, SUITOR, AI speech reorganising. Emotion mouse obtains the
emotional condition like the heart beat, temperature etc. and obtains physiological data with the use of
different sensors. A webcam is used in the magic pointing for determining the pupils of the user under
realistic as well as variable lightning conditions (Bhardwaj et al., 2015). In the case of AI Speech
Reorganition, the user uses the microphone for speaking to a computer and this speech is filtered and
stored it in the RAM. When the user makes eye contact with the SUITOR, the device gets active and can
automatically detect his interest.
The algorithms chosen for detecting emotions are Image Processing Edge Detection Algorithm
improved by Canny Operator, and Rapid Object Detection improved by Viola and Jones. In the edge
detection algorithm, an improved canny detection algorithm is introduced since the traditional algorithm
faced difficulties in treating images. An approach is introduced for minimizing the computation time and
achieving high accuracy in the Object detection or Face detection algorithm (Sen et al., 2018). The
approach creates a face detection system that is 15 times faster as compared to the previous approach.
2

The proposed system depends on the Blue Eyes technology that relates to emotion sensory environment
while the traditional authentication systems used the fingerprints, iris image and thumbprints. There are
various techniques proposed for identifying emotional state and refers to facial affect program for
specifying the relationships between facial movements and specific emotions. The system includes an
interface for determining human gestures and facial emotions.
Application: The media player automatically plays song according to the user’s emotions, acts as a
website plug-in, recommends for YouTube, act as a personal assistant etc.
Hardware and Software Requirements: The hardware requirements for the project include 4GB RAM,
Intel i3, Speaker and Webcam. Software requirements include Python 2.7 and Open CV 3.1.
Functional and Non-functional Requirements: Functional requirements are service statement provided
by the system, and it deals with the system reaction to certain inputs. It identifies and learns the image
captured by the webcam. Machine learning supports vector classification with the help of a vector
machine. Non-functional requirements include the constraints and system properties that arise from
budget constraints, external factors like privacy registrations, organizational policies and user needs.
Progress to date
SVM:
SVM or support vector machines are supervised models that associate with machine learning algorithms
for analyzing the data used in regression analysis. An SVM algorithm creates a model for assigning new
examples to a particular category or other, makes it a linear classifier.
Performance: An SVM model represents examples in such a way that it becomes easy to divide the
examples into separate categories. An SVM constructs a hyper-plane of a hyper-plane set in high-
dimensional space that is used for regression analysis or classification. The hyper-lane having the largest
distance to training data achieve a good separation since with the increase in the margin there is a
reduction in generalization error (Ding et al., 2015). A webcam streams real-time image through a
computer or a computer network. The video stream is saved viewed and shared to other networks
through the systems like the email attachment and internet.
3

Edge detection:
Processing of an image that is considered as a digital image is dealing with the continuation of a PC that
is recognized as the digital computer. This remains a way of utilization regarding the algorithm made
through computers for performing the processing of image upon image that is considered as a digital
image. Corner observation is dealing with a general device that is utilized in order to process a digital
image. Generally, this basic device has been used in order to observation of features moreover removal
that fix its goal as for identifying tiny points within the specific image that is in digital form whereby the
image’s luminousness transformation properly moreover investigate disconnection (Xie and Tu, 2015).
The objective concerning the observation in order to edge remains significant due to reduce the value
concerning the major information within a picture moreover conserve the characteristics of the
framework in order to continue the processing of digital image. Within the image that is recognized as a
grey standard picture, in this case the edge is considered as the general function that, through within
specific nearby different areas within each concerning that a gray standard is much rather low constant
through within several amounts upon the pair of slides concerning a specific and significant edge.
Problems:
There several issues, which a person may face during the utilization of any, edge detector that are
mentioned in the below section of this discussion:
We need to pick edge esteems and the width concerning your veils. In the event that you twofold the
size of a picture leaving its dim qualities unaltered, every one of the inclinations will be divided. The
issue is regular for all inclination based techniques. These confound the setting concerning any edge.
The assessed position concerning any edge ought not for being influenced by the size concerning the
convolution veil (Aslam Khan and Beg, 2015). An extra issue is that the width concerning the cover
moreover consequently the level concerning smoothing influences the places concerning zero
intersections and greatest power slopes in the picture.
This is able for cause significant troubles for line naming plans since they depend on having corners and
intersections being stamped appropriately. Edges are regularly missed in light concerning the fact that
the one dimension slope at corners is generally little. For instance, for discovering bar highlights you
could search for pinnacles from a second subsidiary administrator. Shrewd designed an administrator for
4

discovering lines. First subordinate administrators will just discover step-like highlights. On the off
chance that you need for discovering lines, you need an alternate administrator.
Performance:
1. Possibility concerning fake corners
2. Possibility concerning abstaining corners
3. Failure into an individual calculation concerning the particular angle of the corners
4. Permission over corners those are distorted moreover distinct characteristics for example edges
including unions
Face detection:
The initiate stage in order to recognize the face remains observation of face rather is able to normally be
considered likewise the process that is named as face localization. This is for identifying moreover
localization concerning particular face (Ranjan et al., 2015). The technology that is named as face
observation remains vital due for helping applications for example scanning of lip that is recognized as
an automatic step, consideration of expression related to face moreover observation of individual’s face.
The structure in order to observation of face moreover consideration is mostly likewise.
Generally, this particular structure includes a pair of segment that is considered as functional segments
those utilize in order to detect the image of a particular face as well as due to recognize a specific face.
The detector for images of specific faces investigates in order to human’s faces regarding the picture
moreover localizes the specific and significant human’s faces concerning the past-situation. Thus a
human’s face has been observed or localized, thus, the consideration process has began.
Problems:
There several issues, which a person may face during the utilization of any, face detector that are
mentioned in the below section of this discussion:
1. Existence of basic parts: There might be other extra parts on the face, for example, scenes,
moustache or facial hair. These parts may have extraordinary types, shapes, hues and surfaces.
5

2. Appearance of Face: The outward appearance takes after legitimately on the individual's face.
3. Blockage: Another person or something might mostly discourage a face when the picture is
caught among groups (Chen et al., 2016).
4. Picture Direction: It includes with the variety in pivot of the camera's optical hub.
5. Thinking Condition: The state of a picture relies upon the lighting and camera attributes.
Performance:
There are several features which the face detector provides, has mentioned in the below section of the
discussion:
1. It provides best capability moreover best kind concerning the set of information.
2. It provides several expressions that are related to face.
3. It provides better forbearance of brightness.
4. It provides several types of major poses.
Planned work
Steps in SVM:
The purpose of SVM is finding a hyper-plane in N-dimensional space, which classifies the distinct data
points. For separating two data point classes, there are various possible hyper-planes. The objective of
the project is finding a plane with a maximum margin that is the maximum distance between both data
point classes. Some reinforcement is provided by increasing the margin to the maximum for classifying
the future data points.
Hyper-planes are defined as decision boundaries used for helping the classification of data points. Data
points fall on one side of hyper-plane that is attributed to various classes (Gepperth and Hammer, 2016).
6

Support vectors are considered as a data point, which influences the orientation and position of hyper-
plane. We can increase the classifier margin to the maximum with the help of support vectors (Kamble
and Kulkarni, 2016). Deletion of support vectors can change the hyper-plane position.
Large Margin Intuition is a logical regression where we squash the output received from linear function
within the range [0, 1] with the help of sigmoid function. In case, the squash value is more as compared
to the threshold value it is assigned to label 1, and if the squash value is less than the threshold value, it
is assigned to label 0.
Steps in edge detection:
1. Detection: figure out which edge pixels ought to be disposed of as commotion and which ought
to be held.
2. Localization: decide the precise area of an edge. Edge diminishing and connecting are normally
required in this progression.
3. Enhancement: apply a channel to improve the nature of the edges in the picture (honing).
4. Smoothing: smother however much clamor as could be expected, without devastating the
genuine edges.
Steps in face detection:
This process involves four major stages, which are mentioned in, the below section of this discussion:
1. The first step is to detect the face
2. After that the normalization process has been done.
3. The up next step is to extract the functionality.
4. The final step is consideration of the face, which has been detected
7

Conclusion and Recommendations
The emotion detection media player is utilized for automating and offering a better music experience to
the user. The application resolves the basic music needs of listeners without creating troubles. It utilizes
technology for increasing system interaction in several ways. It makes the user’s work easy by capturing
an image, observing speech, determining emotion and suggesting a customized playlist through an
advanced system. The project includes three algorithms: edge detection algorithm, face detection
algorithm and SVM algorithm. As per the proposed work, the edge detection algorithm is the best
algorithm used for emotion based media player.
8

Bibliography
Aircconline.com. (2019). [online] Available at: http://aircconline.com/sipij/V4N3/4313sipij06.pdf
[Accessed 10 Jul. 2019].
Alamareen, A., Al-Jarrah, O. and Aljarrah, I.A., 2016. Image mosaicing using binary edge detection
algorithm in a cloud-computing environment. International Journal of Information Technology and Web
Engineering (IJITWE), 11(3), pp.1-14.
Aslam, A., Khan, E. and Beg, M.S., 2015. Improved edge detection algorithm for brain tumor
segmentation. Procedia Computer Science, 58, pp.430-437.
Bhardwaj, A., Gupta, A., Jain, P., Rani, A. and Yadav, J., 2015, February. Classification of human
emotions from EEG signals using SVM and LDA Classifiers. In 2015 2nd International Conference on
Signal Processing and Integrated Networks (SPIN) (pp. 180-185). IEEE.
Chen, D., Hua, G., Wen, F. and Sun, J., 2016, October. Supervised transformer network for efficient
face detection. In European Conference on Computer Vision (pp. 122-138). Springer, Cham.
Ding, S., Zhao, H., Zhang, Y., Xu, X. and Nie, R., 2015. Extreme learning machine: algorithm, theory
and applications. Artificial Intelligence Review, 44(1), pp.103-115.
Gepperth, A. and Hammer, B., 2016. Incremental learning algorithms and applications.
Kabani, H., Khan, S., Khan, O. and Tadvi, S., 2015. Emotion based music player. International Journal
of Engineering Research and General Science, 3(1), pp.2091-2730.
Kamble, S.G. and Kulkarni, A.H., 2016, September. Facial expression based music player. In 2016
International Conference on Advances in Computing, Communications and Informatics (ICACCI) (pp.
561-566). IEEE.
Ranjan, R., Patel, V.M. and Chellappa, R., 2015, September. A deep pyramid deformable part model for
face detection. In 2015 IEEE 7th international conference on biometrics theory, applications and
systems (BTAS) (pp. 1-8). IEEE.
9

Sen, A., Popat, D., Shah, H., Kuwor, P. and Johri, E., 2018. Music playlist generation using facial
expression analysis and task extraction. In Intelligent Communication and Computational
Technologies (pp. 129-139). Springer, Singapore.
Sun, X., Wu, P. and Hoi, S.C., 2018. Face detection using deep learning: An improved faster RCNN
approach. Neurocomputing, 299, pp.42-50.
Xie, S. and Tu, Z., 2015. Holistically-nested edge detection. In Proceedings of the IEEE international
conference on computer vision (pp. 1395-1403).
10

Appendices:
Appendix1: Hyper-plane vector
(Source: Bhardwaj et al., 2015)
Appendix2: Support Vector
11