Abstract 1: Dr. M Sanjay Assistant Professor Dept. of Electrical Eng NIT Calicut Kerala India 673601 Email: msanjay@nitc.ac.in Chapter 12 BRAIN COMPUTER INTERFACE MuhamedJishad T K1 and M Sanjay1 1 Department of Electrical Engg, National Institute of Technology Calicut Kerala India ABSTRACT Brain Computer Interface (BCI) is a system of software and hardware that establishes a direct connection between brain and external devices by making use of signals acquired from the brain. In principle, any type of brain imaging techniques such as Electroencephalography (EEG), Magnetoencephalography (MEG), Functional magnetic resonance imaging (fMRI), Functional near-infrared imaging (fNIR) etc. can be used to deviseBCI systems. However, the neuroelectric signals such as EEG are the widely studied signals for BCI applications due to their simplicity, low cost and improved response time when compared to other signal types.The invasive recording of neuroelectric signals like Electrocorticography (ECoG) and Intracortical Neuron Recording (INR) can provide better quality signals than scalp recorded EEG, but with a cost of greater risk to the individual. Additionally, in hybrid BCI systems various other bio signals such as EMG, ECG etc. are also utilized in addition to the signals directly acquired from the brain in order to enhance the performance of the system. Signal acquisition, feature extraction, feature translation, and output device are the primary components of a BCI system.In general, a BCI system detects and extracts features of the brain signal in order to identify the user’s intentions and translate it into useful commands to devices that carries out the desired task. BCI systems find applications mainly in clinical field to rehabilitate people with neuromuscular diseases. Also, efforts are being made to integrate BCI with multimedia, entertainment and gaming applications.Even though there are many successful BCI applications developed by several research groups, this field is still in its infant stage and continuing effort is required to make the systemmore practical for human use. The purpose of this chapter is to give an overview of various types of BCI systems, their working and applications by giving emphasis to signal acquisition, feature extraction and classification methods. The chapter beginswith an introduction to various neuroimaging techniques that are relevant to BCI applications. Even though the BCI applications of fMRI, fNRI, MEG etc. are reviewed in this chapter, the main focus will be on EEG based BCIs which are more popular among researchers due to the simplicity of application. The chapter further explores theBCI systems based on different modalities of EEG such as Slow Cortical Potential (SCP), Sensorimotor Rhythms (SMR), P300 event related potentials and Steady State Visual Evoked Potentials (SSVEPs)with the help of existing literature. Techniques to improve signal quality and feature extraction methods are alsoreviewed in this chapter. Then commonly used data classifiers or classification algorithmsare discussed, which is another important element of BCI that determines the user’s intention by classifying the features extracted from signals acquired.The chapter further looks in to the possibilities of hybrid BCI systems.Case studies of significant works and milestones in BCI research over the years are presented in between the topicsfor better understanding.The chapter is concluded with an analysis of the state of the art and future trends inBCI research. ********** Note: interested for more chapters Abstract 2: Dr. K. A. Venkatesh, Professor Myanmar Institute of Information Technology(MIIT) Mandalay, Myanmar Email: ka_venkatesh@miit.edu.mm> Efficient classification algorithms for Bio signal processing using EEG Sherly Maria Research Scholar Dept. of Computer Science Christ(Deemed to be University) Dr. Chandra J. Assistant Professor Dept. of Computer Science Christ(Deemed to be University) sherly.maria@res.christuniversity.inchandra.j@christuniversity.in Key Words: EEG, Stress, Signal Processing, Feature Extraction, Classification. EEG has been widely used to understand brain signals mainly for its non-invasive method and high accuracy. Signal processing techniques have been used to understand what these signals mean and to predict occurrence of diseases and in return could be prevented. The aim of this chapter is to discuss various signal processing techniques for EEG signals in order to interpret the EEG readings. Stress has been prevalent in humans since a long time. With increasing complexity in our work life, the stress is only rising exponentially. Since a lot of light is now being shed on it, there is a lot of studies going on studying the causes of stress and how to reduce it. EEG is one of the noninvasive methods to record brain waves and one of the most accurate. In order to understand what the signals mean; classification has to be done. Classification plays a major role in understanding what the waves mean, it helps preventing potential diseases from occurring by detecting the diseases like epilepsy and Alzheimer’s. In order to classify these signals effectively, it is important to select a proper classification algorithm based on the data available. There are different methods of classification that could be used, like parametric, non-parametric, time-frequency methods. This chapter would mainly focus on why signal processing is required and most commonly used signal processing techniques for classification. The basic understanding of these models and an explanation of how it is used in signal processing. Among classification, the most commonly used classifiers like Linear discriminant analysis, SVM, ANN, Deep learning would be covered in detail with their uses, advantages and disadvantages. This would help researchers understand the classifiers and models better and would enable them to choose the right model or classifier for their research accordingly based on the requirements and the kind of data they use. Signal processing in EEG has always been a complex task. It is however very important for the researchers to be able to choose the right tool that will enable them to come to the right conclusion. Abstract 3, 4 and 5 Dr. K. A. Venkatesh, Professor Myanmar Institute of Information Technology (MIIT) Mandalay, Myanmar Email: ka_venkatesh@miit.edu.mm> Chapter Proposal for Emerging Trends in Brain and Behavioral Science Proposed Chapter: Auditory Processing &Computational Brain Author: Aravind Kumar Rajasekaran, Associate Professor,Department of Speech Pathology & Audiology,National Institute of Mental Health and Neurosciences, Bangalore and K A Venkatesh Professor MIIT, Mandalay, Myanmar(Adjunct Professor, IIIT BANGALORE, INDIA). Abstract:Speech the verbal mode of communication sets humans well apart from other beings and surely formed the strong basis on which today’s civilizationis built upon. Development of speech (output) wholly depend on the hearing (input). Hearing (Auditory) domain performs a very challenging task of processing of the input sound, as the stimuli per se is transient and dynamic. Cochlea the sensory end organ of hearing feeds the electrical information to the auditory nerve, which in turn connect to the central auditory pathway reaching the cortex. The inputs undergo transformations at multiple levels of the brainstem & and the cortex for various outcomes satisfying specific purposes. The auditory pathway terminates at the temporal lobes, which is rightly termed us the ‘computational hub’. The temporal lobe employs efficient computational mechanisms to decode the signal for its physical properties (frequency & intensity) in time domain and then transfers the output to other related brain areas for the final interpretation and storage. This chapter has six sections. The first section introduces the concept of the Central auditory pathway & and central auditory processing, the second section discusses various methods to study auditory processing, section three deals with Brain stem & auditory processing, section four talks on Auditory cortical processing, section five deal withNeural Networks and computational models in auditory processing and final section seven is on Brain computer interface- Deciphering the input before output. Keywords:Auditory processing, auditory neuroscience, computational models, brain computer interface. Approximate pages : 25-38 Abstract 4: Chapter Proposal for Brain and Behavior Computing Proposed Chapter:Role of Statistics in Brain Computing and Behavioral Science Author:K A Venkatesh Professor MIIT, Mandalay, Myanmar(Adjunct Professor, IIIT BANGALORE, INDIA). Abstract: A Brain Computer Interface (BCI) also known as Brain Machine interface. BMI, is a collection of hardware bundled with software systems that enables human to interact with their environment using control signals generated by electroencephalographic activity. In BCI, the most important task is feature selection and selection from the six popular neuroimaging system, irrespective of exogenous, endogenous, Synchronous and Asynchronous BCI. Every thinking process result in an different patterns of the brain signals. Hence, BCI can be viewed as pattern recognition system. Brain signals will have different information in each activities. Different channels capture these signals. The foremost aspect of feature extraction is dimensions reduction then classification and finally statistical analysis. This chapter focuses on statistical analysis and it contains five sections. First section is to on Null Hypothesis testing, section 2 is on two groups t test, section 3 is on one way anova, section 4 is on correlation and regression, section 5 is on factorial anova. Keywords: Testing of Hypothesis, correlation and regression, anova Approximate pages: 25-30 Abstract 5: Chapter Proposal for Brain and Behavior Computing Proposed Chapter:Role of Statistics in Brain Computing and Behavioral Science Author:K A Venkatesh Professor MIIT, Mandalay, Myanmar(Adjunct Professor, IIIT BANGALORE, INDIA). Abstract: A Brain Computer Interface (BCI) also known as Brain Machine interface. BMI, is a collection of hardware bundled with software systems that enables human to interact with their environment using control signals generated by electroencephalographic activity. In BCI, the most important task is feature selection and selection from the six popular neuroimaging system, irrespective of exogenous, endogenous, Synchronous and Asynchronous BCI. Every thinking process result in an different patterns of the brain signals. Hence, BCI can be viewed as pattern recognition system. Brain signals will have different information in each activities. Different channels capture these signals. The foremost aspect of feature extraction is dimensions reduction then classification and finally statistical analysis. This chapter focuses on statistical analysis and it contains five sections. First section is to on Null Hypothesis testing, section 2 is on two groups t test, section 3 is on one way anova, section 4 is on correlation and regression, section 5 is on factorial anova. Keywords: Testing of Hypothesis, correlation and regression, anova Abstract 6: Chapter Proposal for Brain and Behavior Computing Proposed Chapter: Introduction and Data Acquisition Author: K. A. Venkatesh Professor MIIT, Myanmar and Anandi Giridharan Principal Research Scientist, Indian Institute of Science, Bangalore Abstract: Enormous increase in big data, learning analytics, and Edge Intelligence presents new technique to redesign user centric adaptive learning. In this chapter we are proposing to design next generation adaptive learning systems based on new developments in learning and technology. Outline of the chapter: This chapter has five sections: first section briefly discusses about Informal framework of Adaptive learning architecture. The second chapter focuses on proper use of data acquisition methods in adaptive teaching environment. The third, sections presents techniques of user centric knowledge construction, Section four shows that students find adaptive learning contents for students better learning strategies. Key words: Data acquisition, Adaptive learning, edge intelligence, Learner model Abstract 7: Dr.K C Raveendranathan <indran@ieee.org> Chapter 4 Data Science in Brain Signal Computing Abstract The model of the human brainas a massive, Artificial Neural Network (ANN) is quite popular among several researchers and in many disciplines of science. It is apparent to note that the human brain is a massive data processor which involves many signals-biological, and external data inputs from several sources. It is estimated that an average the human brain contains 86 Billion (86×1012) neurons and this depicts the complexity of the system. It is interesting that the biosignals emanating from the brain, the Electro-Encephalo Gram (EEG) signals,are of the order of 2µV to 100µV at 0.5Hz to 100Hz. In contrast to this, there also exists the evoked potential of the human brain in response to external visual and auditory stimuli which are termed as Evoked Potential/Event-Related Potential (EP/RP). These are usually of the order of 0.1µV to 20µV at a frequency of 1Hz to 3kHz. We in this chapter attempts to study in details the processing of Brain Signals in the context of brain and Human-ComputerInterface. We show that the signals emanate from the brain are tantamount to typical Data Science conditions, by way of its statistical nature and volume. For medical applications needle electrodes and implanted electrodes are used to sense the EEG. On the other hand, applications involving Human-Computer Interface (HCI) use surface electrodes only. One of the major problems associated with surface electrodes is that the measured signal will be a superposition of several signals from many neurons. Usually, the processing of brain signals involves preprocessing to eliminate the noise and artefacts present in it. Digital processing of EEG signals also involves the usage of suitable Digital-to-Analogue Converters (ADCs) and Digital Signal Processors. One has to use intelligent signal processing algorithms to analyse the EEG signals properly. It is well known that different brain regions are used for performing different tasks or functions such as psychomotor functions, emotional engagements, sensory perceptions, and visual processing. Thus, the measured signal from the skull will reflectthese spatially distributed functions of the brain, and always a combined signal. Over the past two decades, the signal processing hardware and software has matured and improved immensely and become very inexpensive. Very effective, computationally intensive signal processing algorithms can now be implemented on Field Programmable Gate Arrays (FPGAs) and low-priced signal processors. A recent innovation in technology is the biologically inspired computing, which intent to mimic the functioning of the human brain. To design and develop systems that are true substitutes to the brain, one has to do pattern recognition from the sensed signal and data. The foremost step to pattern recognition is feature extraction. The extracted features of the signals are used to train a model data set either in a supervised or unsupervised fashion. Thus we have two models of machine learning-the unsupervised and the supervised. To effective control the learning rate, one has to resort to feature selection techniques as part of pattern recognition. Thus it is imperative that the brain signal processing involved principles of Data Science as well. Abstract 8: Dr.Arun Sasidharan1, KusumikaKrori Dutta2 arun.s@axxonet.net, kusumika@msrit.edu 1Senior 2Assistant Research officer & Research Program Head,Axxonet Brain Research Laboratory (ABRL), Bengaluru Professor, Department of electrical and Electronics Engg, M.S.Ramaiah Institute of Technology, Bengaluru. MACHINE LEARNING TECHNIQUES FOR BRAIN AND BEHAVIOUR MEASUREMENTS USING ELECTROENCEPHALOGRAPHY Abstract Brain is often considered a highly advanced prediction machine that accomplishes this enormous feat through complex interaction with its billions of cells called neurons. Due to their inherent electrical nature, such neuronal activities appear as blurred electrical field changes over scalp. These can be measured non-invasively as voltage changes using extremely sensitive instruments called Electroencephalography (EEG). Thus, EEG patterns capture many aspects of mental processes like cognition, behaviour, and emotions, across the illness to wellness spectrum. Though several years of training would enable clinicians to visually interpret important features of EEG signals, these are extremely cumbersome, and the majority of the patterns get discarded as noise for simplicity. Over the years, several signal processing techniques have helped to extract many such EEG features that were invisible to even the most experienced eyes. This has allowed researchers to expand the applications of EEG into many more areas, which may be broadly categorized into Brain decoding (like detecting behaviour and emotion) and Anomaly detection (like identifying Alzheimer’s disease, Seizure, Sleep stages). However, the above applications can involve enormous amounts of EEG features, making the task overtly daunting. This is where machine learning (ML) techniques have become crucial. These are basically algorithms involving complex mathematical concepts and programming, which performs a specified function with the given data and progressively gets better over time. Thus complex tasks can become efficient and scalable through automation. This allows considerable improvement in EEG interpretation in terms of accuracy and interpretability for clinical use, as well as usability for online applications like brain computer interface (BCI). There are several ML techniques (like K-nearest neighbour, support vector machine, random forest, etc.) as well as their more evolved form called deep learning techniques (like convolutional neural networks, recurrent neural networks, etc.) that are commonly used by researchers for more accurate interpretation of EEG signals, despite their random, non-linear, non-periodic and noise-prone nature. This chapter will initially provide an overview to the various ML and deep learning algorithms for multiclass, time series classification from EEG data. Subsequently, it will guide the readers through the application of ML, using seizure detection and sleep stage detection as examples. The chapter will stress on how to select EEG features based on neurobiological grounding and how to decide on an appropriate ML pipeline. Though the main focus of this chapter is ML application on EEG signals, the same concepts will be applicable in several other time series data (like fMRI BOLD signals, audio signals, etc.). Abstract 9: An efficient single trial classification approach for Devanagari script based P300 speller using knowledge distillation and transfer learning Ghanahshyam B. Kshirsagar, Narendra D. Londhe* Department of Electrical Engineering, National Institute of Technology Raipur, Chhattisgarh, India, 492010. *corresponding author- nlondhe.ele@nitrr.ac.in Abstract: Background: In existing Devanagari script based P300 speller (DS-P3S), deep CNN and weighted ensemble of deep CNN had implemented to improve the performance of classification of P300 from multiple and single trial respectively. However, DCNN that too its ensemble variant becomes cumbersome and increases the computational burden of a system due to large number of trainable parameters. Hence, a spatial-temporal based CNN had implemented to reduce this computational burden. However, the compressing the deep model also leads to loss of information which further costs accuracy. Method: Therefore, to compensate this loss, we are proposing first time the utilization of knowledge distillation to implement a compressed CNN with comparable performance. The proposed work includes following contribution:1) an efficient compact model for single trial P300 detection, 2) channel-wise convolution to provide sparse connectivity, 3) knowledge distillation to compensate the loss and reduce the trade-off between accuracy and number of trainable parameters, and 4) transfer learning to handle intertrial and inter-subject variability. For experimentation, we have used a self-generated DS dataset which includes ERP-EEG of 79 characters acquired from 10 health subjects. Results: An average accuracy of 90.82%±1.47 and 88.74%±2.08 are achieved for cross and within subject classification in single trial using proposed compact model. There is a positive increment of 24% in average accuracy with reduction of trainable parameters by 3-54 times than existing DCNN, shallow and compact CNN. This remarkable performance ensures the effectiveness of knowledge distillation and transfer learning in the implementation of a compact classification model for DS-P. Abstract 10: Md. Hedayetul Islam Shovona*, D (Nanda) Nandagopal ,Jia Tina Du , and Vijayalakshmi Ramasamya,b a a b a, School of Information Technology and Mathematical Sciences, University of South Australia, Adelaide, Australia shomy004@mymail.unisa.edu.au Miami University, Oxford, Ohio 45056, United States of America *Corresponding author Directed Functional Brain Networks: Charactisation of information flow direction during cognitive function using Non-linear Granger Causality Human Information search and information retrieval is a cognitive function which applies some level of cognitive load on the user. With the phenomenal growth in affordable mobile devices the need for complex information in day to day life has been increasing significantly. This has an impact not only on the user’s information search behavior but also on the cognitive load. The measurement of dynamic and global interactions of the brain regions during cognitive activity is a real challenge. Therefore, in this chapter we examine the dynamic behavior of functional brain networks during both visual and web search tasks. The impact of information search and retrieval on cognitive function using EEG based functional brain networks is studied. Very few previous studies have considered the changing dynamics of information flow direction in the human functional brain network during cognitive function. This novel study is aiming to explore the information flow direction during cognitive activities. Direct measure of brain activity during information search is an interesting but challenging topic for information retrieval researchers. Directed functional brain networks were constructed using nonlinear Granger causality (GC) from multichannel Electro Encephalogram (EEG) data collected during eyes open ( baseline), visual search and Web search tasks. The constructed functional brain networks were analyzed using complex network metrics including connectivity density, clustering coefficient and local information flow measures as well as our proposed weighted Information Flow Direction Pattern algorithm. The empirical analysis suggests that increased cognitive activity takes place in the occipital and frontal lobes during visual and Web search phases compared to baseline state. The significant increase in both the connectivity densities and clustering coefficients during the search tasks clearly demonstrate higher cognitive activity in the brain demonstrating increased cognitive load. The information flow patterns together with the sensitivity of the complex network metrics to cognitive states observed in this study have potential applications in examining the effects of cognitive abilities on information search processes. The results of quantitative measures of cognitive load during both visual and web search tasks demonstrated in this chapter may contribute towards designing better Information search and retrieval systems leading to least cognitive/mental load on the user and thus improving the search and retrievalbehaviors’. Keywords: Granger Causality, information flow, directed functional brain network, EEG, cognitive activity, visual search, web search. Abstract 11: From : Dr. Dr. Mohan Awasthy Principal G H Raisoni Institute of Engineering and Technology Nagpur BE Electronics, M. Tech Computer Technology Ph.D. Electronics and Telecommunication Sr. Member of IEEE Email: awasthymohan@gmail.com Study of Neuro Fuzzy System for Human Emotion This Chapter explain the rational structure and field of study of cognitive intelligent system (CIS) or Affective Computing and its importance on Emotional State Recognition. In recent years study of Emotional Intelligence has become a widespread research domain. Several research has been conducted to create machines that can sense and understand human affective states, such as emotions, behavior, mood and the interests as well. It tries to fill the communication gap between human and computers with the sense of feelings. The work is based on a comprehensive literature survey and referential study of Emotional Intelligence which illustrates its connection with emotional states recognition domain Abstract 12: Dr. Rahul Upadhyay Post-Doctoral, Trinity College Dublin, Ireland Assistant Professor Electronics and Communication Engineering Department Thapar Institute of Engineering & Technology (Deemed to be University) Patiala, Punjab India Mobile No.: +91-9755130755 E-mail: rahul.upadhyay@thapar.edu, rupadhya@tcd.ie Abstract 13: Komal Jindal, Rahul Upadhyay, Hari Shankar Singh Department of Electronics and Communication Engineering, Thapar Institute of Engineering and Technology, Patiala 147 004, Punjab, India Email: komaljindal24@gmail.com, rahul.upadhyay@thapar.edu, harishankar1990@gmail.com Application of TQWT-Entropy based Automated Epileptic Seizure Identification An epileptic seizure is a group of disorders caused by a synchronous aberrant discharge of neurons in the human brain. Early epileptic seizure detection using a Computer-Aided Diagnosis scheme may help in improving the physiological condition of patients. In this work, an Electroencephalogram (EEG) activity-based automated epileptic seizure detection scheme is proposed using appropriate Tunable QWavelet Transform (TQWT)-Entropy features. The TQWT can reliably represent the significant arrangement of constraint time-frequency subbands from considered EEG activity. The proposed methodology begins with the decomposition of given EEG activity into six time-frequency sub-bands using TQWT. In the next step, four Entropy features are extracted from decomposed time-frequency sub-bands. For classification of the extracted features, three machine learning procedures i.e., SVM, RF and NN are employed. Classification results are evaluated corresponding to extracted feature vectors, and it is observed that classification accuracy of 100% (using SVM), 99.66% (using ANN) and 98.33% (using RF) is attained using only single Entropy feature. It is observed that the proposed TQWT-Entropy based epileptic seizure detection methodology shows significant classification results in terms of classification accuracy, sensitivity, and specificity. Among all other classifier algorithms, the SVM classifier emerges as the most suitable classifier algorithm in present work. Abstract 14: Evolutionary Comparison with Performance Analysis of Metaheuristic Optimal Controller for Arc Load Rajkumar Jhapte1, Anish Pratap Vishwakarma2 1Department 2Department of EEE, SSTC, SSGI, Bhilai, India of EE, National Institute of Technology, Meghalaya, India 1jhapte02@gmail.com 2anishvishwakarma9766@gmail.com Abstract. In recent decades, the randomly variable loads such as Speed variant drives, Modern Power electronics loads, automated machineries and some high transient current carrying loads such as Arc furnace etc are proved as a responsible factor for increase in Harmonics and cause of circuit parameter disturbance results in declining the power quality of system. With this regards has approach towards minimization of disturbing parameter and to enhance the performance of Optimization Technique for controller based filters is done here. Various novel optimal control techniques like Genetic Algorithm (GA), Particle swarm optimization (PSO), Gravitational Search Algorithm (GSA), Harmony Search Algorithm (HSA) and Cuckoo Search Algorithm (CSA) for tuning the parameter of PI Controller are utilized throughout these work to approaching towards system stability. The main focus of the work is to extract the reference current after acquiring the load current to the controller which is in our case taken as a non linear current from an arc furnace load. The control strategies for evaluate reference current is Synchronous Reference frame (SRF) Theory among various topology available. This industrial load is taken into consideration only after the relevant V-I characteristics supports the real time operating status of Electric Arc Furnace Load (Chaotic model simulated). The proposed system is implemented with adoptability and in contrast the employability of metaheuristic optimization techniques are being observed and simulated to meet the actual disturbing effects of non linear loads as Electric Arc furnace. Also the ISE, IAE and ITAE criteria are have been implemented to refine the absolute error so that the optimization itself being more deliberate towards the harmonic minimization. Moreover the reduced percentage of THD is obtained with ITAE criteria. All the result are simulated in MATLAB 2017a Simulink environment. Chapter 15: Monali Gulhane Research Scholar KL University,Vijayawada,India Asst Prof JIT,Nagpur ,Maharashtra,India monali.gulhane4@gmail.com 9766709211/9075089800 Chapter No 10 Machine Learning Techniques for Brain and Behavior Computing Author1: Monali Gulhane Author2: T.Sajana Abstract Modern understanding of human actions is focused primarily on universal characteristics of visual images. In recent decades, conventional human activity identification has been largely focused on the global characteristics of an image, introducing various static features such as edge features, shape features, statistical features or transforming features to characterize human behavior. This human behavior are examined with maximum accuracy in the area of Machine Learning. Machine learning is an application of artificial intelligence (AI) that provides syste ms intelligent to automatically learn and develop from experience of data received and process without being any specifically programmed .This application of artificial intelligence (AI) plays an vital role in human behavior prediction .Hence the chapter provides brief knowledge about the techniques of brain computing and human behavior analysis through machine learning. The learning cycle begins with experiences or evidence, such as facts, direct knowl edge, or feedback, to search for correlations in the data and make informed choices in the future based on the facts we have. The primary objective is to allow comput ers to automatically learn without human involvement or assistance, and to modify behavior accordingly. There are numerous techniques to examine the brain activity and human behavior in machine learning but in this chapter we will focus on deep learning techniques for brain computing because deep learning has efficient advantages for brain interface and analysis . Many BCI (Brain Computer Interfaces) have implemented deep learning techniques for brain computing due to its advantages that its not time consuming in preprocessing with all feature engineering steps i.e. it starts working directly on raw data signals of brain to learn the distinguish information of the back propagation data. For human behavior computing analyzing the motions or actions of all body parameter(hands,eyes,legs,heads etc) are important in human behavior prediction or computing hence for this purpose also deep learning engine of machine learning is in boom of Artificial Intelligence. In earlier day the storage of data in computer and performance of computer system restricted the advancement of deep learning in bioparameter analysis but now a day the use of high end computer system and cloud are helping to store large amount of data which inherited the development of deep learning in human behavior computing and analysis . The chapter also gives and idea of type of deep models called as Deep Neural Networks (DNNs). According to the research DNNs can directly work on raw inputs to automate the functions . The chapter also describes the user behavior using actions, activities, and intra- and inter-activity behavior. Researcher Aitor Almeida and Gorka Azkune, have created a deep learning architecture based on long short-term memory networks (LSTMs) that models the inter-activity behavior. This chapter will give you brief of machine learning usage ,techniques for brain and human behavior computing with its merits and demerits along with brief knowledge of human behavior prediction algorithm. Chapter Flow: 1. Introduction to Machine Learning(Definition Approaches ) (700 words) 2. Machine Learning Algorithm for Brain Computing(Introduction Paragraph) (1500 words) 2.1 Working of Brain and its data computing process 2.2 Brain Interface Computing:(Definition, Classification) 2.3 Power and Limitations of Algorithm for Brain Computing 3. Machine Learning Algorithm for Human Behavior Computing (Introduction Paragraph) (1500 words) 3.1 Elements of User Behavior 3.2 Deep neural networks with models