Heritage Institute of Technology

About: Heritage Institute of Technology is a based out in . It is known for research contribution in the topics: Steganography & Support vector machine. The organization has 581 authors who have published 1045 publications receiving 8345 citations.

Buckling characteristics of cut-out borne composite stiffened hyperbolic paraboloid shell panel:

Abstract: The present study investigates buckling characteristics of cut-out borne stiffened hyperbolic paraboloid shell panel made of laminated composites using finite element analysis to evaluate the gover...

Hypothesizing the applicability of the principle of linear combination in predicting sensing behaviors of quantum dots: A deeper understanding of the precise tuning of quantum dot properties with capping composition

Abstract: In this paper we intend to demonstrate, for the first time, the relevance of the hypothesis based on the linear combination principle (LCP) in devising materials that may exhibit intermediate of their pure-state-like behaviors. To the best of our knowledge, the present study embodies the first attempt in demonstrating why the hypothesis can be pertinent for nano-sized materials. In doing so, we have picked up the CdS QDs as model system and two structurally similar thiols as capping agents (3-mercaptopropionic acid, MPA, and thiolactic acid, TLA). Here, we attempted fabrication of mixed dual thiol capped CdS QDs (abbreviated as mDTCQ) employing mixture of two thiols to assess the effects of composition of capping mixture on the sensochemical characteristics of QDs. We showcased that the selectivity and also sensitivity of mDTCQ toward two metals (Ni and Pb) via photoluminescence (PL) turn-off based strategy can be tuned as per linear combination principle which is found to hold good as a function of mole percentages of capping agents used during synthesis. Importantly, in the present context; we would like to draw the attention of the readers that the results are the stepping stones toward understanding the origin of senso-selection of nanomaterials and is not just something that can be misinterpreted as incremental to the existing reports.

An efficient power delivering scheme for sensorless drive of Brushless DC motor

Abstract: This paper presents a photovoltaic (PV) panel fed sensorless Brushless DC motor (BLDC) drive using Sheppard –Taylor (S–T) converter for load matching in order to make the system energy efficient. In this work, a detailed study on S–T converter is done as it helps to stabilize the fluctuating output voltage of standard PV panel. Hence, in this application model, we have considered S–T converter as a feed to the six switch three phase Brushless DC (BLDC) motor drive in which the rotor position is estimated using back electromotive force (EMF) detection technique. This work is developed and studied on MATLAB/Simulink platform as well as in real time implemented hardware. The simulation model has borrowed the motor parameters (viz. stator back-EMF, rotor RPM and max. output torque, etc.) and circuit components from the implemented hardware. Finally, a detailed study of the implemented S–T converter reveals a significant performance improvement over a conventional Buck–Boost converter for same PV panel. The output results show that the scheme described in this scope of work efficiently handles the DC power generated by solar cell using S–T converter as the rotor RPM offset is negligible in closed loop motor drive.

Gait signal classification tool utilizing Hilbert transform based feature extraction and logistic regression based classification

Abstract: In this paper, we have employed a machine learning approach for automatic classification of healthy and pathological gait signals and subsequent identification of the neurological disorder in the pathological gait signals. The machine learning algorithm we have proposed is the Logit model of the Logical Regression Classifier. As the process of walking is automatically controlled by the nervous system it is important to develop a non-invasive method so that patients with serious neurological disorders like Huntington's disease and Parkinson's disease receive early medical attention and they get proper care before they are more affected. Swing, Stance and double support intervals (expressed as percentages of stride) of 63 subjects were analyzed. In this paper, a relevant gait signal feature extractor is developed which is combined with Logistic Regression Classifier to classify healthy subjects and pathological subjects. Analysis of real-time gait signals is simplified using the Hilbert Transform which converts the real signals into an analytic signal. The proposed algorithm was developed using the MATLAB platform and the average accuracy of multiclass classification is found to be 86.05% while the accuracy of detecting healthy subjects from pathological subjects is 87.79% and the accuracy of classifying subjects having the Huntington's disease and Parkinson's disease is found to be 85.22%.

Scalable Recommendations Using Clustering Based Collaborative Filtering

Abstract: The development and growth in recommender systems address the issue of information overload faced by the online users while searching for products or services. However, recommender systems typically face challenges like data sparsity and scalability as they often handle large datasets. The most widely used recommendation technique is Collaborative Filtering (CF) that pins down the recommendations on the opinions of the most similar users. The core of a CF algorithm is the similarity computations among the users or items, which becomes extremely expensive when new users and items join the system at a very rapid rate. The proposed work deals with this scalability problem by implementing a clustering based CF approach. Typically in a recommendation problem there exists a set of users, a set of items and a rating matrix, that records the ratings assigned by the users to the items. In this work, we first partition the set of users using a CURE (Clustering using representatives) based method and then leverage the resultant clusters to formulate recommendations for the target user. In the proposed method, the CF algorithm is not applied to the entire user-item database, rather the algorithm is applied separately to each of the clusters resulting in reduced recommendation time. Moreover, Clustering also helps to improve the sparsity problem by reducing the dimension of the rating matrix and filtering out noisy data. The results of the experiments conducted on MovieLens-10M and MovieLens-20M datasets indicate that our method significantly reduces the runtime and at the same time preserves good recommendation quality.