Dr. Madhu B R
@jyothyit.ac.in
Professor and Head, Department of Artificial Intelligence & Machine Learning
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Pooja S, Prasanna Kumar K, A S Sushmitha Urs, Vaibhavi B Raj, Madhu B R, and Vinod Kumar S
IEEE
Amidst the backdrop of technological convergence, this study delves into the realm of augmented visual intelligence. It does so by orchestrating a harmonious fusion of the YOLOv3 framework, deep learning paradigms, and the seamless integration of Text-to-Speech (TTS) capabilities. The research paper meticulously dissects a methodological blueprint designed for the creation of an object recognition system, impeccably tailored to the precise identification of objects within the comprehensive COCO dataset. With the YOLOv3 architecture as our cornerstone, meticulous parameter fine-tuning transpires within the Darknet framework, ensuring an unswerving alignment with the diverse object categories that define the COCO dataset. Our system demonstrates proficiency by deploying TTS technology to deliver real-time auditory interpretations of recognized objects, enhancing both user accessibility and engagement. The ethical compass steadfastly guides our approach, encompassing privacy safeguards that underscore our commitment to the conscientious and responsible utilization of data. System performance is rigorously assessed through the lens of pivotal metrics, including precision, recall, and the F1 score, validating the system’s precision and reliability. This research elucidates the transformative potential innate to the amalgamation of deep learning and TTS integration within the sphere of object recognition, thus carving a path for pioneering applications and the evolution of technology.
N. Shravan, M. Manoj Kumar, Sriraag Jayanth, R. S. Bindu, B. R. Madhu, and K. S. Sreekeshava
Springer Nature Singapore
A S Sushmitha Urs, Vaibhavi B Raj, Pooja S, Prasanna Kumar K, Madhu B R, and Vinod Kumar S
IEEE
This research intends to build an effective and quick algorithm for identifying the alphabets in American Sign Language (ASL) using natural hand movements, increasing communication accessibility for people with hearing impaired limitations. The system's ultimate goal is to act as a translator between spoken language and sign language, enabling more effective and efficient communication between those with hearing loss and others who don't have any hearing loss. The research uses image processing, machine learning, and CNN-based artificial intelligence to recognize ASL movements and generate outputs that are simple to interpret. The potential impact of this work on communication accessibility for people who have hearing loss is significant.
Madhu B. R., Dinesh M. N., Tsewang Thinlas, and Deril Menezes
International Journal of Electrical and Computer Engineering Institute of Advanced Engineering and Science
Power quality is the most significant factor of power sector. The end user equipment such as induction motor, inverters, rectifiers inject harmonics into power system that influences the quality of power delivered. The presence of harmonics forces the use of instantaneous reactive power theory to calculate instantaneous power that helps in finding the compensating currents to eliminate harmonics. The control action required by active filter is accomplished by STM32F303RET6 microcontroller. Single phase induction motor is used as a dynamic nonlinear load in one of the three phases and resistive loads on the other two phases. TRIAC based RC triggering circuit was used to control the single phase induction motor. This paper presents the simulation and hardware implementation of shunt active power filter for 3 phase 4 wire unbalanced system. The hardware results show that THD in the source side has been reduced from 50.7% to 9.6% by implementing the SAPF.
Madhu B R, Dinesh MN, and Ravitheja BM
Institute of Advanced Engineering and Science
<p>The quality of power is a major concern because of harmonics in the lines due to various sources. Passive filters have been used to eliminate the specific harmonics. In this paper shunt hybrid active power filter which consists of shunt active filter and shunt passive filter is designed to reduce harmonics in source side for non linear loads. The shunt passive filter includes a tuned RLC circuit that behaves like a band-pass filter for a particular harmonic content. The shunt active filter is a VSI that produces reverse harmonic current based on Clarke’s transformation. This paper presents the control strategy by implementing instantaneous reactive power theory to design Shunt Hybrid Active Filter (SHAPF) for a non linear load. The proposed control technique is simulated using MATLAB SIMULINK, Results prove there is considerable reduction in THD of source current from 30.35% to 3.25% with the proposed controller on shunt hybrid active power filter. The performance of SHAPF is better than Passive and Shunt Active Power Filters.</p><p> </p>
B. R. Madhu and M.N. Dinesh
IEEE
A three phase inverter, controlled by SPWM technique driving a three phase induction motor with low pass passive filter is implemented on MATLAB simulink platform. The harmonic content present in the three phase inverter is analyzed using FFT tool. Hardware implementation of the same is done using DSP (TMS320F2812) controller. The output from the inverter is connected to a three phase induction motor and the speed control is carried out by constant Voltage/frequency method on both open loop and closed loop. The FLUKE 434 series II power analyzer is used for recording the total harmonic distortion (THD) produced at the inverter output. The three phase, 1H.P, 415V, 50Hz star connected induction motor is used for this purpose. In the closed loop system PI controller is connected in the feedback. The experimental result shows that the induction motor speed can be controlled by implementing SPWM technique. The experimental results of the induction motor drive were validated with the theoretical results with a good speed regulation and the desired performance of the drive system has been achieved. The THD content was reduced from 63.4% to 3.44% using low pass passive filter which is well below 5% as per IEEE standards.
Anku Jaiswal, A. S. Manjunatha, B.R. Madhu, and Murthy P. Chidananda
IEEE
Intrusion is one of the most serious problems with network Security, as new types of intrusions are getting much more challenging to detect. Large amount of network traffic has been generated due to the use of internet; most of the generated traffic is in the format which cannot be used directly to arrive at meaningful information. The cleansing and labeling of data each time needs a considerable amount of human effort, and is time consuming. In this paper we show how, Semi supervised machine learning technique can be used in intrusion detection, for both labeled and unlabeled data. In the proposed technique we take a small amount of labeled data to create model and using this model we show how to predict the unlabeled traffic. Machine Learning tool is used for this purpose which uses semi-supervised classifier to build the model. The created model is then integrated in Pentaho which with the help of Weka Scoring provides the expected output. The proposed technique helps the network administrator to take quick decision by classifying the incoming traffic as either malicious or normal and hence efficient detection of intrusion.
Prakash Chandra, A. S. Manjunatha, P. Chidananda Murthy, and B. R. Madhu
IEEE
The features like on-demand self service, rapid elasticity, measured service, broad network access has increased the popularity of cloud computing and has motivated the business organization to adopt the cloud as part of their IT services and solutions. Various statistics and survey has proved that the user-base of the cloud is increasing day by day, which has enforced the cloud service providers to ensure the high availability and reliability of their services to maintain good credibility in the market. The cloud is powered by the virtualization technology, which makes it possible to deliver the computing services in the form of Virtual Machines (VMs). The efficiency of any cloud system greatly depends on its VM allocation policy, which is the policy adopted by cloud service provider to allocate the virtual machine to the available physical server. This paper presents the proposed technique of using genetic algorithm to find the optimal solution to allocate and map the virtual machines to the hosts and hence maximizing the utilization of resources. The experimental analysis carried out on CloudSim tool has proved that the proposed technique is more efficient and can provide good scope for future research.
Madhu B.R., Manjunatha A.S., Prakash Chandra, and Chidananda Murthy P
ENGG Journals Publications
Cloud service providers are experiencing the huge demand for the computational power as most of the companies are outsourcing their IT services to the cloud. This increased popularity of cloud computing has resulted in the establishment of large-scale datacenters with the huge number of computational intensive servers. These datacenters are very much expensive to maintain as they consume enormous amount of electrical energy, which results in the emission of green house gases like carbon dioxide. One of the main reasons for the huge power consumption is the inefficiency of the existing resource scheduling algorithm, which utilizes the existing resources in the non-power aware manner. This paper is based on the idea that the power consumption is directly proportional to the CPU utilization of the server, and hence proposes the task consolidation technique which maps the users requests i.e. task to the virtual machine in such a way that it maximizes the CPU utilization and minimizes the energy consumption at the same time. CloudSim is used as the simulation tool kit to show how the proposed task scheduling technique is more power efficient compared to the existing task scheduling technique. KeywordCloud computing, Task consolidation, CloudSim, Energy consumption, Green computing
Chidananda Murthy P., Manjunatha A.S., Anku Jaiswal, and Madhu B.R.
ENGG Journals Publications
One of the most difficult and unsolved issues in network is the security issue, because of continuous evolving nature of both threats and the measures used to detect and avoid threats. Among different types of attacks, one of the most vulnerable attacks in network security are bots that consume the resources maliciously and exhaust them. Malicious Cloud Bandwidth Consumption (MCBC) attack is a new type of attack, where the aim of the attacker is to consume the bandwidth maliciously, in turn causing the financial burden to the cloud service host. MCBC is generally vulnerable to the internet based web services in public cloud. MCBC mainly aims at frequently consuming the bandwidth in a slow manner, hence affecting the pay-as-you-go utility model, causing the consumer in the form of monetary loss. Unlike DDOS attack which is short lived and makes the resource unavailable to the user, MCBC attack is a long term attack which slowly attacks the target for an extended period and remains undetectable. As this attack does not affect the availability issue immediately, it is not discussed much as DDOS attack. This paper discuss about how machine learning technique can be used to detect the MCBC attack in the form of request per second, any traffic violating this range are classified as MCBC attack. The proposed system consists of using semi supervised machine learning which uses labeled network traffic for building model and unlabeled traffic to classify using the built model. KeywordNetwork Security, Machine learning, MCBC attack, Supervised learning
Madhu B.R and M.N. Dinesh
IEEE
This paper presents miniature hardware model of 6 pulse controlled rectifier circuit for a monopolar HVDC link. The Controlled rectifier is interfaced with PIC (16F72) microcontroller for generating various firing angles. The results of output voltage are recorded using digital oscilloscope and are compared with the theoretical values at different firing angles. Performance analysis parameters such as voltage ripple, ripple percentage, harmonics with and without low pass filter for R and RL load is obtained by varying the firing angle. Results show that the ripple factor and ripple percentage has drastically reduced with the addition of the low pass filter and also smoothens the DC output voltage which were observed in digital oscilloscope. There has been a reduction of 92% ripple for R load and 60% for RL load (for firing angle α= 90°).
P. Jain Pratik and B. R. Madhu
IEEE
Data mining has been gaining popularity in knowledge discovery field. In recent years, data mining based intrusion detection systems (IDSs) have demonstrated high accuracy, good generalization to novel types of intrusion, and robust behavior in a changing environment. Still, significant challenges exist in design and implementation of production quality IDSs. Masquerade attacks pose a serious threat for cloud system due to the massive amount of resource of these systems. This paper presents a Cloud Intrusion Detection System (CIDS) for CIDD dataset, which contains the complete audit parameters that help in detecting more than hundred instances of attacks and masquerades that exist in CIDD. It also offers numerous advantages in terms of alert infrastructure, security, scalability, reliability and also has data analysis tools.