Delhi Technological University

About: Delhi Technological University is a education organization based out in New Delhi, India. It is known for research contribution in the topics: Computer science & Control theory. The organization has 4427 authors who have published 6761 publications receiving 71035 citations. The organization is also known as: Delhi College of Engineering & DTU.

New scheme for underwater acoustically wireless transmission using direct sequence code division multiple access in MIMO systems

Abstract: This paper proposes a new technique based on Direct Sequence Code Division Multiple Access for underwater acoustically wireless transmission with excessive transmission rate. Environment of subsea is challenging for wireless communication because the medium in which waves are propagating is not air. In fact, it is propagated through fractions of water having different densities. Finding out various techniques for multipath access targeting the physical layer of Acoustic Sensor Networks is indeed necessary. The recent approaches have suggested that coded modulation techniques with exploited diversity are highly preferred in order to enhance the dependability of the acoustic link in different multipath channels. The proposed technique divides the channel into sub ones and transmits information via those sub channels. In variety-spectrum, a signal in a bandwidth is unfold within frequency domain and broad bandwidth. Experimental results show that Bit Error Rate (BER) of this method is better than that of channel equalization in the respective systems.

Mixed-Mode Universal Biquad Filter Using OTAs

Abstract: This paper introduces an electronically tunable mixed-mode universal biquad filter configuration employing four single output operational transconductance amplifiers (OTAs), one dual output OTA and...

Reliability modeling using Particle Swarm Optimization

Abstract: Software quality includes many attributes including reliability of a software. Prediction of reliability of a software in early phases of software development will enable software practitioners in developing robust and fault tolerant systems. The purpose of this paper is to predict software reliability, by estimating the parameters of Software Reliability Growth Models (SRGMs). SRGMs are the mathematical models which generally reflect the properties of the process of fault detection during testing. Particle Swarm Optimization (PSO) has been applied to several optimization problems and has showed good performance. PSO is a popular machine learning algorithm under the category of Swarm Intelligence. PSO is an evolutionary algorithm like Genetic Algorithm (GA). In this paper we propose the use of PSO algorithm to the SRGM parameter estimation problem, and then compare the results with those of GA. The results are validated using data obtained from 16 projects. The results obtained from PSO have high predictive ability which is reflected by low error predictions. The results obtained using PSO are better than those obtained from GA. Hence, PSO may be used to estimate SRGM parameters.

Multi-input integrative learning using deep neural networks and transfer learning for cyberbullying detection in real-time code-mix data

Abstract: Automatic detection of cyberbullying in social media content is a natural language understanding and generic text classification task. The cultural diversities, country-specific trending topics hash-tags on social media, the unconventional use of typographical resources such as capitals, punctuation, emojis and easy availability of native language keyboards add to the variety and volume of user-generated content compounding the linguistic challenges. This research focuses on cyberbullying detection in the code-mix data, specifically the Hinglish, which refers to the juxtaposition of words from the Hindi and English languages. We explore the problem of cyberbullying prediction and propose MIIL-DNN, a multi-input integrative learning model based on deep neural networks. MIIL-DNN combines information from three sub-networks to detect and classify bully content in real-time code-mix data. It takes three inputs, namely English language features, Hindi language features (transliterated Hindi converted to the Hindi language) and typographic features, which are learned separately using sub-networks (capsule network for English, bi-LSTM for Hindi and MLP for typographic). These are then combined into one unified representation to be used as the input for a final regression output with linear activation. The advantage of using this model-level multi-lingual fusion is that it operates with the unique distribution of each input type without increasing the dimensionality of the input space. The robustness of the technique is validated on two datasets created by scraping data from the popular social networking sites, namely Twitter and Facebook. Experimental evaluation reveals that MIIL-DNN achieves superlative performance in terms of AUC-ROC curve on both the datasets.