Hitesh Shah

Physical activities recognition from ambulatory ECG signals using neuro-fuzzy classifiers and support vector machines

Abstract The use of wearable recorders for long-term monitoring of physiological parameters has increased in the last few years. The ambulatory electrocardiogram (A-ECG) signals of five healthy subjects with four body movements or physical activities (PA)—left arm up down, right arm up down, waist twisting and walking—have been recorded using a wearable ECG recorder. The classification of these four PAs has been performed using neuro-fuzzy classifier (NFC) and support vector machines (SVM). The PA classification is based on the distinct, time-frequency features of the extracted motion artifacts contained in recorded A-ECG signals. The motion artifacts in A-ECG signals have been separated first by the discrete wavelet transform (DWT) and the time–frequency features of these motion artifacts have then been extracted using the Gabor transform. The Gabor energy feature vectors have been fed to the NFC and SVM classifiers. Both the classifiers have achieved a PA classification accuracy of over 95% for all subjects.

Automatic thresholding in edge detection using fuzzy approach

Edge detection is important step in image processing. In spite of two decade of research the need for general purpose edge detector is still felt. Threshold decision is the key uncertainty in the edge detection algorithms. Soft computing approach represents a good mathematical framework to deal with uncertainty of information. In this work, we used fuzzy logic for automatic thresholding and generated threshold is used with Sobel methods for edge detection. Empirical performance study shows the efficacy of automatic threshold method.

Edge detection techniques using fuzzy thresholding

Edge detection is forefront of image processing algorithms. There are problems associated with different edge detectors. Thresholding is critical problem in edge detection. The paper represents three different methods of fuzzy based thresholding and resulted adaptive threshold is used in edge detection algorithm. Simulation results shows fuzzy thresholding based edge detection gives better results than conventional methods.

A REINFORCEMENT LEARNING ALGORITHM WITH EVOLVING FUZZY NEURAL NETWORKS

Abstract The synergy of the two paradigms, neural network and fuzzy inference system, has given rise to rapidly emerging filed, neuro-fuzzy systems. Evolving neuro-fuzzy systems are intended to use online learning to extract knowledge from data and perform a high-level adaptation of the network structure. We explore the potential of evolving neuro-fuzzy systems in reinforcement learning (RL) applications. In this paper, a novel on-line sequential learning evolving neuro-fuzzy model design for RL is proposed. We develop a dynamic evolving fuzzy neural network (DENFIS) function approximation approach to RL systems. Potential of this approach is demonstrated through a case study–-two-link robot manipulator. Simulation results have demonstrated that the proposed approach performs well in reinforcement learning problems.

Protein Secondary Structure Prediction Using Support Vector Machines (SVMs)

Bioinformatics or computational biology is field of science in which biology, computer science and information technology merges into a single discipline. In modern computation biology, protein secondary structure prediction is a major problem. Secondary structure prediction is depends on its amino acid sequence. Current studies prefer machine learning techniques for classification and regression task. Recently many researchers used various data mining and machine learning tool for protein structure prediction. Our intention is to use model based (i.e., supervised learning) approach for protein secondary structure prediction and our objective is to enhance the prediction of 2D protein structure problem using advance machine learning techniques like, linear and non-linear support vector machine with different kernel functions. The datasets used for this problem are Protein Data Bank (PDB) sets, which is based on structural classification of protein (SCOP), RS126 and CB513.

Face Recognition Using 2DPCA and ANFIS Classifier

With the growth of information technology coupled with the need for high security, the application of biometric as identification and recognition process has received special attention. The biometric authentication systems are gaining importance, and in particular, face biometric is more preferred for person authentication because of its easy and non-intrusive method during acquisition procedure. Face recognition is considered to be one of the most reliable biometric, when security issues are taken into concern. Various methods are used for face recognition. To recognize the face, feature extraction becomes a critical problem. In this paper, two-dimensional principle component analysis (2D-PCA) has been applied for feature extraction. The feature vectors are then applied to adaptive neuro-fuzzy inference system (ANFIS) classifier. The result indicates that ANFIS classifier yields 97.1 % of classification accuracy.

3D Face Recognition Based on Pose Correction Using Euler Angle Method

Face recognition is one of the biometric method, to identification of given face image using main features of the face. 3D face recognition approach handles the challenges of 2D face recognition such as pose, illumination, expression, etc in a better way. The exploration of 3D face recognition is widely used for security at many places like airport, organizations, crime detection etc. uncontrolled condition of real-world biometric applications is pose which is a great challenge to any face recognition approach. Such pose variations can cause extensive occlusions, resulting in missing data. This work presents 3D face recognition method which is invariant to pose. The method handles pose correction problem with rotation matrix based on Euler angle. Further dimensionality of face image is reduced by the principal component analysis and the recognition is done by the Euclidean distance algorithm. we used GAVAB 3D face database for simulation and measured performance like Recognition rate, False acceptance rate (FAR), False rejection rate (FRR) and Equal error rate (EER) are used to evaluate performance of the method. We have achieved 1.12 % improvement in Recognition rate and 0.15% improvement in equal error rate for probes with neutral and non-neutral, respectively.

An anomaly detection in smart cities modeled as wireless sensor network

Smart city is an important application of the recent technology — Internet of Things (IoT). IoT enables wide range of physical objects and environments to be monitored in fine detail by using low cost, low power sensing and communication technologies. While there has been growing interest in the IoT for smart cities, there have been few systematic studies that can demonstrate weather practical insights can be extracted from the real time IoT data using advanced data analytics techniques such as anomaly detection. We carried out a case study of smart environment based on real time data collected by the city of Aarhus, Denmark. We analyzed and find the levels of different air pollution elements to detect the unhealthy or anomalous locations based on Air Quality Index (AQI). Machine learning framework namely neural network, Neuro-fuzzy method and Support Vector Machines for both binary and multi class problems has been used for anomalous location detection form pollution database. Simulation results using MATLAB show that Machine learning techniques are reliable in terms of accuracy and calculation time for smart environment.

Energy efficient link adaptation using machine learning techniques for wireless OFDM

Energy Efficiency in wireless communication is very important due to the slow progress in battery technology with improvement in technology. In this paper, we applied energy efficient link adaptation using Machine learning techniques in Orthogonal Frequency Division Multiplexing (OFDM). We sounding the channel condition periodically and observing the channel parameters. Our aim is to select the optimal mode of the channel which maximizes energy efficiency or throughput of data subject to a given quality of service (QoS) constraint. Simulation results show that the proposed solution achieves significant improvement over existing link adaptation algorithms. Presented work aims on maximizing the throughput and provides orders of magnitude gain in energy efficiency linked to poorly chosen fixed modes when used for energy efficiency maximization purposes.

Genetic algorithm for energy harvesting-wireless sensor networks

Traditional routing protocols such as LEACH, PEGASIS, TEEN etc. are no longer appropriate for the Energy Harvesting-Wireless Sensing element Networks (EH-WSN). Requirement is that WSN must have low energy consumption. Machine Learning algorithms can be used for minimizing energy consumption. Thus our main objective is to develop machine learning based routing protocol, which is having energy harvested from environment instead of batteries. Research work done on LEACH and modified LEACH algorithms in order to achieve energy efficiency in WSN. Majorly these approach concentrate on clustering of sensor nodes and/or modifying the routing protocol. Finally we compared our proposed machine learning algorithm with LEACH to show energy efficient network with improved network lifetime.