Pardhu Thottempudi
@bvrithyderabad.edu.in
Assistant Professor, Department of Electronics and Communications Engineering
BVRIT HYDERABAD College Of Engineering For Women
Pardhu Thottempudi became a member (M) of IEEE in 2015. Pardhu was born in Luxettipet village in Adilabad district in Telangana state, India. He completed Batchelor’s degree B.tech in the stream of electronics and communication engineering in 2011 from MLR Institute of Technology, Hyderabad, India. He has done his master’s degree M.Tech in embedded systems from Vignan’s University, Vadlamudi in 2013. He is pursuing Ph.D in the stream of RADAR signal processing from VIT University His Research Includes Human Motion Analysis Behind walls using Optimized Deep Learning Algorithms. His major fields of interests include Digital Signal Processing, RADAR communications, embedded systems, and implementation of signal processing on applications in FPGA. He is working as assistant professor of department of Electronics and Communication Engineering in BVRIT HYDERABAD College of Engineering for Women, Hyderabad, India since 2023.
EDUCATION
VIT University, VELLORE- Thesis Submitted (2023)
M.Tech- Vignan University, Vadlamudi- 2013
B.Tech- MLRIT, Hyderabad- 2011
Intermediate-2011
Tenth-2005
RESEARCH, TEACHING, or OTHER INTERESTS
Signal Processing, Artificial Intelligence, Communication
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
Thottempudi Pardhu and Vijay Kumar
Springer Science and Business Media LLC
Pardhu Thottempudi
IGI Global
Machine learning (ML) is a powerful tool that unveils hidden insights from internet of things (IoT) data. These technologies enhance decision-making in education, security, business, and healthcare. In healthcare, they automate tasks such as maintaining records, predicting diagnoses, and monitoring patients in real time. However, different ML algorithms perform differently on various datasets, influencing results and clinical decisions. Understanding these ML algorithms and their application in handling IoT data in healthcare is crucial. This chapter highlights key ML algorithms for classification and prediction, providing an in-depth overview of their role in analyzing IoT medical data. The analysis reveals that different ML prediction algorithms have unique limitations, necessitating careful selection based on the dataset type for accurate healthcare predictions. The chapter also illustrates the use of IoT and ML in predicting future healthcare trends.
Pardhu Thottempudi
IGI Global
This chapter delves into the development of a machine-based project aimed at detecting and identifying human faces, a process known as face recognition. This process not only discerns human faces but also determines whether the face is familiar or unfamiliar, with advanced iterations even providing the identified individual's name. OpenCV, a tool within the realm of image processing, is utilized to ascertain the detected face. The implementation of this method occurs in two phases: training and testing. The project's design primarily incorporates live-stream face detection, feature extraction, and recognition of detected faces from a stored database. This technology has potential applications in criminal identification and security surveillance in police investigations. The machine-based nature of the project significantly reduces the likelihood of errors compared to manual recognition. The chapter concludes by comparing the detected and processed faces with a database to recognize familiar faces, thereby verifying the individual's real identity.
Pardhu Thottempudi and Nagesh Deevi
IGI Global
The main objective of the project is to implement fingerprint recognition because it is one of the most popular and reliable methods for human identification and because it makes use of minutiae, which are unique features found in fingerprints. The fingerprint is another type of biometric that is employed to recognize individuals and verify their identities. Extraction of information from fingerprint scans is among the most important steps in fingerprint recognition and classification. The proposed approach for the project relies on utilizing a variety of methods and algorithms to identify fingerprints using the ROI method (i.e., threshold & centroid algorithm). Two human fingerprints can be compared using ROI to determine which has more detail. The main method for highlighting the minute details of the sample fingerprint's fingerprint is FFT extraction. A percentage score is produced as a result of the minute data, and it indicates whether or not two fingerprints match. It was written in MATLAB code.
Pardhu Thottempudi, Nagesh Deevi, Amy Prasanna T., Srinivasarao N., and Mahesh Babu Katta
IGI Global
One of the many applications of machine learning in healthcare is the analysis of large amounts of data to reveal new therapeutic insights. Once doctors have this data, they can better serve their patients. Therefore, satisfaction can be raised by using deep learning to enhance the quality of care provided. This work aims to integrate machine learning and AI in healthcare into a single system. Predictive algorithms based on machine learning could revolutionize healthcare by allowing doctors to avoid unnecessary treatments. Various libraries, including those for machine learning algorithms, were used to develop this work. Because of its extensive library and user-friendliness, Python has emerged as the preferred language. syntax. The authors used various classification techniques to train machine learning models and then select the one that provided the best balance between accuracy and precision while avoiding prediction error and autocorrelation problems, the two main causes of bias and variance.
D. V. S. Chandrababu, Pardhu Thottempudi, Ch. Babaiah, and G. Koushik
AIP Publishing
D. V. S. Chandrababu, Pardhu Thottempudi, Ch. Babaiah, and G. Koushik
AIP Publishing
Sreedhar Kollem, Katta Ramalinga Reddy, Sreejith S, Ch Rajendra Prasad, Srinivas Samala, and Thottempudi Pardhu
IEEE
Image distortion may result from a variety of factors, such as changes in electronic imaging equipment that create noise. The goal of blur image alignment is to determine which images are blurred and to restore them. Due to this, a unique blurred image restoration approach is developed, which consists of a point spread function, canny edge detection, GLCM extraction, and general regression neural network for identifying the type of blurred images, and a wiener filter for restoring the image. This technique uses a combination of the General Regression Neural Network (GRNN) and Deep Neural Network (DNN) to detect the kind of a blur and determine the calibration efficiency of the DNN and the degradation efficiency of the GRNN. The mean square error (MSE), the covariance factor, and the peak signal-to-noise ratio (PSNR) are used to assess the effectiveness of the proposed method. In comparison to conventional procedures, the proposed method yields superior outcomes.
Pardhu Thottempudi, Venkata Surya Chandra Babu Dasari, and Venkata Surya Prasad Sista
Springer Singapore
Thottempudi Pardhu and Vijay Kumar
International Information and Engineering Technology Association
Now a day’s defence applications associated to novel, army and military war fields are required wall imaging discrimination. As of now many wall-imaging techniques are designed but cannot discriminate the target and clutter with accurate working. Therefore, a novel advance wall image tracking method is required for differentiate the clutter and human target. In this research work single value decomposition technique is used to estimate the range bin behind the wall target. In order to track the target and clutter single-value-decomposition (SVD) is not sufficient, so that along this SVD, threshold skewness (TS) method has been presented. Combination of SVD-TS giving the accurate long range-bin sensing and directed the human’s targets. SVD-TS method is a statistical scheme, which can realise the amplitude ranges through large number of range-bin scans. This technique improves the accuracy by 98.6%, skewness by 8%, and normalised power by 98.9%. These SVD-TS method is more efficient and compete with existed techniques.
K Jyothi, Thottempudi Pardhu, R Karthik, and T S Arulananth
IEEE
This paper presents a methodology to measure the level of focus and to identify the areas of the skin with high levels of Specular-reflection that are uploaded to the BTBP mobile application for skin analysis. Measuring this information makes it possible to give the user feedback on the quality of their image capture, and how to improve future captures. Specular-Reflection is excessive reflection from the skin's surface has a detrimental effect on skin analysis. In order to avoid such problems we need to identify areas of intense Specular-Reflection and eliminate those areas from analysis. The concept used here is histogram analysis. Each time an image is captured, the camera's auto-focus is used to try and achieve optimal focus. Occasionally this can fail and result in an image that is out-of-focus and therefore undesirable In order to reject these images, focus level quantification is necessary. Referring to the edge information of the objects in the image, calculate the level of sharpness in the image. Filtering and image enhancement concepts will be used for this implementation. This will be implemented using the C# platform and will utilize the Open CV library. This includes learning about Image declarations, image data reading and assigning in C#.
Thottempudi Pardhu and Bhaskara Rao Perli
IEEE
Watermarking is a technology that complements cryptography by embedding imperceptible signals in work. Information hiding techniques have recently become important in a number of application areas. Digital audio, video and images are increasingly furnished with distinguishing but imperceptible marks, which may contain a hidden copyright notice that helps to prevent unauthorized copying directly. One approach is to introduce an invisible image known as digital watermark, into an image sequence. This paper presents secured algorithms for embedding digital watermarks into images. The proposed method performs imperceptible watermarking of images in frequency domain. The watermark is embedded in the DWT and DCT domain of an image in a multi-resolution way. In the decoding phase, once the watermark is extracted from the watermarked image, certain performance measures such as peak signal to mean noise ratio (PSNR) and correlation are calculated. Different types of attacks have been applied to the watermarked image to test the robustness of the applied technique and for each case, PSNR and correlation are calculated. Even though DWT method gives better results when compared to DCT, performances of both methods are good.
Thottempudi Pardhu and N.Alekhya Reddy
IEEE
In this paper Low power multipliers can be designed by using new technique. This technique introduces new hybrid full adders and compressors. The new adders allow NAND gates to generate most of the multiplier partial product bits instead of AND gates and inverters which were used in signed multipliers like Baugh-wooley multipliers, thereby lowering the power consumption and the total number of required transistors. In this paper 8×8 array multipliers, Baugh-Wooley multipliers are designed by using this hybrid adders and tree multipliers are designed by using compressors and the power, area, delay of these multipliers are compared. For an 8×8 implementation, the ALL-NAND array multiplier achieves reduction interms of power consumption and 7.8 % reduction in transistor count with increase in time delay compared to baugh-wooley multiplier. The ALL-NAND tree multiplier exhibits lower power consumption and 7.3% reduction in transistor count, with longer time delay, compared to conventional tree multiplier. The simulation results are obtained by tsmc 0.18nm technology.
Thottempudi Pardhu and Vijay Kumar
Inderscience Publishers
TWI technology is an emerging research area used in urban warfare situation. Ultra wide band range signals has the capability of penetrating through the materials hence ultra wideband RADAR is suitable for this situation. To give the better representation of the scanned scene signal processing techniques are included on the raw data. In this paper, we used impulse RADAR for TWI and for the retrieval of B-scan signal back projection algorithm is used. Here we used a gain factor of 100 to increase the strength of echo. To reduce the clutter from B-scan signal SVD technique is used and the peak SNR is observed to be −3.30 dB.
Thottempudi Pardhu and Sunkara Harshitha
IEEE
The main aim of this paper is to design the digital frequency meter with frequency analyzing module. As a measuring frequency's instrument, the cymometer often refer to as electronic counter in modern electronic technology. Its basic function is to measure the signal frequency and this frequency counter has a wide range of applications. It is not only used in the general simple instrument's measurement, but also used in teaching, research, high-precision instruments measuring, industrial control and other areas. Currently, the high-performance and simple-structure electronic products have become the mainstay of market. In this project the cymometer not only measure the frequency but also determines the jitter, glitches status etc., Digital Frequency Meter were designed using VHDL language and simulated using Modelsim 6.3v simulator. In order to increase the efficiency of the system, sleep mode concept is introduced, which reduces unwanted circuit power utilization.
Thottempudi Pardhu, S. Manusha, and Katakam Sirisha
IEEE
In this project, a Flash ADC with TIQ(Threshold Inverter Quantizer) comparator and a Wallace tree encoder is described. Both DC and transient analysis report for 3 bit ADC are included. A comparison between Wallace tree and ROM encoder is included in the project illustrating how power can be reduced using Wallace tree encoder instead of ROM encoder. The design was successfully simulated for piece-wise linear and sinusoidal input. The 4 bit and 5 bit flash ADC with TIQ comparator and Wallace tree encoder is also described.
Thottempudi Pardhu, Usha Rani Nelakuditi, and Suresh Pampana
IEEE
In general many real time problems are represented with random variables. But a random variable is characterized by a Probability Density Function(PDF). As per the Central Limit Theorem any stochastic process handled with many random variables can be converged as a normal distribution. Hence normal distribution plays a key role in modeling real time problems contains many random variables. This paper dealt with the generation of Gaussian PDF using uniform random sequence. The generated PDF is validated in LABVIEW.
RECENT SCHOLAR PUBLICATIONS
MOST CITED SCHOLAR PUBLICATIONS
RESEARCH OUTPUTS (PATENTS, SOFTWARE, PUBLICATIONS, PRODUCTS)
1.Power Efficient Compressor Using Full Adder Circuit
Inventor: Thottempudi Pardhu
Status: Published on 29/08/2014 pp:60
Application Number:3975/CHE/2014
2. WEED IDENTIFYING ROVER
Inventor: Thottempudi Pardhu
Status: Issued06/10/2021
Design Number: 347292-001
3.ARTIFICIAL INTELLIGENCE BASED HUMANOID ROBOT FOR SURVILLANCE AND SECURITY
Inventor: Thottempudi Pardhu
Status: Case is Amended with Controller
Application Number:377792-001
4.VARIABLE RATING ACCUMULATOR CHARGING STATION WITH TOOLS BOX
Inventor: Thottempudi Pardhu
Status: Granted
Design Number:6270282 (UK Design Patent)
5.DESIGN OF SENTRY ROBOT FOR SURVEILLANCE AND SECURITY
Inventor: Thottempudi Pardhu
Status: Granted
Design Number:6272417 (UK Design Patent)