1. Study of Assessment of Cognitive Ability
G R Sinha, PhD
IEEE Senior Member, ACM Distinguished Speaker, IEEE Distinguished Speaker
Professor, Myanmar Institute of Information Technology Mandalay
Recipient of ISTE National Award, TCS Award, IEI Award, Expert Engineer Award, Young Engineer Award, Young Scientist Award
2. Videos and Motivation
Cognitive Assessment
Background Research
CNN and Sample Result
Findings
2Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Lecture Outline
3. Videos:Neurons-to_Networks.mp4
As per Canadian Institute for Advanced Research Experience and Brain Development: “Stimuli in early
life switch on genetic pathways that differentiate neuron function – sensitive period” “stimuli affect
the formation of the connections (synapses among the billions of neurons)”. This was concluded from
a research based studies in humans, monkeys and rats.
3Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Motivation
4. There is association between IQ and brain size in humans which remains robust across age,
intelligence domain, and sex of participants.
In several literature, brain size is considered as one of many neuronal factors associated with individual
differences in intelligence.
The association in the context of human cognitive evolution and species differences in brain size and
cognitive ability, is not warranted to interpret brain size as an isomorphic proxy of human intelligence
differences.
Reference: Jakob Pietschniga, Lars Penked, Jelte M. Wichertse, Michael Zeilerb and Martin Voracek (2015), “Brain
Volume and IQ”, Meta-analysis of associations between human brain volume and intelligence differences: How strong
are they and what do they mean?
4Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Motivation (contd..)
5. 5Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Perception-Cognition-Action
6. This is a technique that can assess mental abilities, also a formal assessment of an individual’s abilities
in a range of areas, such as verbal and non-verbal skills, memory and speed of processing.
Individuals are asked to do a number of tasks, including puzzles; answering questions or remembering
certain things.
Cognitive assessment can also be using physiological signals such as EEG (Electroencephalography),
ECG (Electrocardiography), GSR (galvanic skin response) etc.
For example, each band of EEG corresponds to a property of human brain and hence can be used to
assess various cognitive abilities of an individual.
References: “Understanding Cognition’, available at http://www.psychologytoday.com/basics/cognition” & “UlricNeisser, Cognitive Psychology,
Prentice Hall, 1967”
6Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Cognitive Assessment
7. Biometrics is common application of image processing and computer vision. The features are trained using
suitable soft computing technique such as neural network.
The neural network architecture may have a number of hidden layers, one input layer and an output layer.
Convolutional neural network (CNN) is a simple example of deep learning network for face recognition.
The faces of different age group, gender, occupation and structure are subjected to learning method using
neural network. Appropriate set of features are extracted and classifiers are used to discriminate the faces
as per their occupation; structure whether the face is inverted or not; gender, etc.
Reference: G. R. Sinha, “Study of Assessment of Cognitive Ability of Human Brain using Deep Learning”, Int. J. Inf. Tech. (Springer), 1(1), pp. 1-6,
2017.
7Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Introduction
8. Marzi and Viggiano employed event related potentials (ERPs) for investigation of face recognition whether
and when brain activity varies as per processing level in the encoding stage of face recognition. The
recognition was assessed by using shallow learning (orientation decision) and deep learning (occupation
decision). Reaction time and correct recognition time were recorded and the learning methods were
compared. Most important finding of the work is the effect of encoding level and deep learning in
recognizing the orientation or occupation of human faces.
Wong and Sun studied the learning models for classification purpose with features extracted in the task.
The deep learning model provides better discriminative and descriptive ability. Regularized deep fisher
mapping (RDFM) was used employing the new learning model and the performance was significantly
improved.
Haque et al. employed deep learning for studying human body and motion dynamics. Rothe et al. suggested a deep learning based solution for
face recognition using CNN architecture of VGG-16 and trained on ImageNet for classification of data. Jain et al.
8Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Background Research
9. Chowdhury presented a study of non-thinking (computer machine thinking) and thinking machines
(human thinking). The article was published in Current science.
Munivenkatappa et al. presented a study of psychological and cognitive domain of human brain very
nicely in traditional way of thinking process. The article attempted to give a direction of research
towards energy transformation in cognitive ability of human brain.
9Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Background Research (contd..)
10. Modelling of human brain is evaluated in terms of its cognitive and discriminative ability. The
experimentation would select hundreds of persons of different age groups and gender; and they will
be shown thousands of different types of images.
During testing process, they will be asked to identify the faces that were shown to them. The reaction
time and the response will be recorded.
Based on the reaction time and response time, the cognitive ability of human brain is evaluated for
different age group and gender.
Analysis varies with different age and possible gender also, whose impact is very important in
understanding the cognitive or recognition capability of the person.
10Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Objectives
11. It consists of following stages:
1. Study of deep learning based research on face recognition
2. Facial image databases: a database of real time faces of thousands of persons is developed. The
faces of persons include faces of all age group, gender and of different occupation.
3. Database of persons required during testing (non-machine learning): another database will include
100 persons or more of different age group and gender and the role of these persons will be in
testing process, wherein they will be asked whether they correctly recognize the faces that were
shown to them. The faces of databases will be shown to all the persons during manual training
process of biometrics.
4. Pre-processing: the faces of databases may have different pose, lighting, and of different size. So,
the normalization all faces is to be done followed by image smoothing or de-noising; if required.
Image de-noising may not be required for non-machine learning but that would be required for
machine learning (deep learning) based face recognition tasks. A framework of image processing
tools for resizing, reformatting and enhancement can be developed for this purpose.
11Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Methodology
12. 5. Feature extraction: features of faces will be extracted that are required in any biometric method. The
features of faces, intensity or gray scale based features are to be extracted using a training method
employing deep learning, which may uses convolution neural network or similar to give better results of
recognition accuracy.
6. CNNs are both fully and locally connected to hidden layers unlike traditional neural network because
for all fully connected networks, the operation becomes computationally intensive. CNNs use parameter
sharing, pooling and dropout also which reduce the number of common features to large extent and
hence addressing the computational issues.
12Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Methodology (contd..)
13. Step 1: Features are extracted by first convolution layer. These features are generally low level features such
as edges and lines
Step 2: High level features are extracted by subsequent layers.
Step 3: Size of input is NxNxD and this has to be convolved with kernels whose size kxkxD separately.
Step 4: Convolution of an input with one kernel produces one output feature, and with H kernels
independently produces H features.
Step 5: Each feature in the output consists of (N - k +1)x(N - k - 1) elements.
Step 6: For each position of the kernel in a sliding window process, kxkxD elements of input and kxkxD
elements of kernel are multiplied and added.
Step 7: The kx kxD multiply-accumulate operations are required for producing one output feature
13Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
CNN Steps
14. The study of cognitive ability of human brain: based on RT, RC and NRC; the cognitive ability is
evaluated using a suitable mathematical function, as the cognitive ability is directly depending on the
assessment measures RT, RC and NRC.
The retention time (RNT) for each person that how long the person remembers the faces that are
shown to him/her.
Cognitive ability is a function of RT, RC, NRC and RNT.
14Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Cognitive Ability
15. 15Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Sample Result
16. For assessment of cognitive ability of human brain, we considered 380 persons (including men and women
of different age group) as recognition agents for recognizing the faces shown to them.
The number of faces in different pose that were shown to all during training their brain to the faces was 25.
This may prove to be path breaking in the research direction leading to study of human brain and its
cognitive capability.
The results also highlight explicit difference between the ability of women and men.
With the age group growing up, the retention time is more and the ability of recognizing and recalling goes
down. Moreover, the time taken to recall is also more.
Women exhibited more cognitive ability than that of men and the interesting fact is that the retention time
was also observed as less for female participants.
16Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Sample Result (contd..)
17. The traditional use of the learning concept has almost been saturated in the field of computer vision and face
recognition.
We suggest deep learning in assessment of cognitive ability of human brain.
Aiming to train thousands of facial images into our image database.
A sample size of 380 persons was tested in real time deep learning based face recognition.
Response time and correct identification were recorded that shows potent research scope of deep learning in
assessment of cognitive ability of human brain at large scale.
The cognitive ability of women was found more than that of women.
However, the sample size of persons’ faces that were shown to 380 persons of different age group and sex could
be increased and the modelling could be developed and compared with deep learning based (machine learning
based) recognition of faces.
17Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018
Findings
18. 18
E= mc2
Thank you, any queries please!
Study of Assessment of Cognitive Ability G R Sinha ACM Distinguished Speaker Lecture July 20, 2018