Hand Gesture based Person Identification has become a promising technology with a wide range of applications for Human Computer Interaction. Hand gestures are recognized by Kinect sensors, Gyro sensors and Inertial Sensors. We have used Frequency Modulated Continuous Wave Radar based hand gestures dataset of 6 people with four different gestures for identifying each Person. Previous works have identified hand gestures with Radar as its sensors, in this work we have proposed a dilated Convolution Neural Network Model for identifying the Person based on micro-Doppler hand gestures using a reduced number of parameters in contrast to existing state of art methods, that work with larger number of parameters for learning. We have observed that the proposed model has achieved an excellent recognition rate of 97.47

As more and more students are seeking counseling services to help them pursue higher education opportunities abroad, it has become apparent that there is a lack of automation and a monopolization of counseling agencies, which presents a challenge for these students. To address this issue, this study has developed HICON AI: Higher Education Counselor. HI-CON made from Higher - Counseling is an innovative application that offers personalized college selection and preparation recommendations to students. By asking a set of defined questions, the bot gets to know the user and provides tailored guidance based on the information provided, utilizing refined Machine Learning models and Retrieval Augmented Generation. This study has specifically used Llama 2, LLM by meta for the considered use case, because of its high performance and financial viability. The developed new product ensures the highest level of accuracy and reliability.

Automatic synthesis of realistic images is complex; even cutting-edge AI and machine learning systems sometimes suffer from not fulfilling this expectation. However, the development of image processing has made it possible to perform operations on an image to improve or extract information from it and synthesize images from textual descriptions, making this a prominent area of study. Based on the idea of text-guided image production, the image inpainting task is replacing the damaged parts of an image with new, appropriate features that make sense in the given context. Researchers in this area have made some encouraging advances using neural image inpainting techniques. In contrast to previously developed text-guided image production methods, inpainting models must first compare the semantic content of the provided text to the remaining portion of the image to determine the semantic information for the missing section. Providing a mask and a text prompt with the information you need to make the desired changes can alter certain parts of an image. To boost the semantic similarity between the produced image and the text, an image-text matching loss is utilised. Further research is needed to resolve the challenge of guessing the missed content with only the context pixels. We propose a computationally efficient system to perform image inpainting on synthetic images that have been generated from textual data while ensuring visual coherence and accuracy.

Brilliant Surveillance utilizing Personal Digital Assistants and smart sensors has become unavoidable in Human Activity Recognition (HAR) because of the development in innovation. Elderly individuals need surveillance through gadgets that ought not be wearable or holdable. COVID-19 has accentuated this requirement for human movement acknowledgment mainly because of the death pace of older people. Radar systems provides solutions for Human activity Recognition without the vibe of being observed with its exceptional highlights as opposed to savvy cameras, has been in dynamic research. In this paper, we have utilized a dataset published by University of Glasgow, in which 56 subject's activities are observed under realistic environments, a first of its kind dataset in the Radar Domain using Frequency Modulated Continuous Wave (FMCW) radar. Using Microdoppler signatures for gait patterns, we have proposed a Convolutional Neural Network (CNN) Architecture model with better optimization, adaptable for real time deployment that classifies all the six distinct activities performed by the subjects under study. It characterizes the spatial information from the spectrograms through layers of the network to identify the distinct patterns. It is shown that the classification model has achieved higher Activity recognition rates with respect to test data indicating good model generalization as well.

The quantity of written content about medical information is staggering, and it keeps on increasing daily. Think about the internet, which is full of websites, news, constantly updated status, blogs, and a great deal more. Utilizing search to swiftly scroll through key findings is the most effective method for navigating material mainly because the results are not arranged in any specific fashion. A considerable percentage of this stuff needs to be summarized to emphasize the most significant information. Creating a custom summary with our own two hands of everything that can be effectively done by text summarization techniques. There is a critical need for this automated method. This project incorporates technologies that are capable of summarizing the text automatically. It implies paraphrasing phrases into a single thought. We propose a hierarchical document that represents an architecture for extraction aggregation that utilizes “Longformer” as a set encoder and “Transformer” as a text encoder to represent text documents properly. This model is referred to as the Long-Trans-Extr model. One of the benefits of using Longformer as a set encoder is that the model can take in lengthy documents with a maximum of 4096 tokens, which only adds a little amount of additional processing. This quick summarization tool can be used by doctors or medical practitioners for quickly analysing emergency or critical patients.

Many colleges employ online learning as media learning where each piece of multimedia, be it sound, film, or text, can receive feedback from students. The lecture wants to know about the emotions that students experienced when they accessed the media, such as happiness, disappointment, or unhappiness, and instructors want to know how joyful they felt. This study built a tool cell for online media to use in identifying emotions in column comments. This paper focuses on user experience (UX) design, which is based on the Human-Centred Design method that prioritizes empathy for users and focuses on user demands to meet the discovery of user wants and develop high-fidelity prototypes. The results of the testing show that the average accuracy is 1.68%, the average recall is 1.55, the average precision is 1.45, and the average accuracy for the use of this phone website is 80% for emotion recognition based only on a column of comment in the internet media.

The emerging growth of artificial intelligence (AI) technology created a way for many innovative developments. Speech to lip synchronization (STLS) is one of the key requirement in applications related to film making, voice modulation, video creation etc. The identification of speaking face synchronized with the voice need to be done accurately to improve the quality of the video. Many existing systems face the problem while imposing the new audio into the existing video files. The movement of lips dynamically changes according to the speaking faces. To solve the missing synchronization problem of existing videos to retrieve the original quality from the video input, the proposed system is focused on creating accurate lip synchronization model using Deep SyncNet (DSN) using Deep learning convolution architecture. The timing accuracy of the video synchronization extensively improve the quality of the video and demands real time experience from the video footage. Detection of facial variations without extracting the features are particularly challenging. The proposed system considers the existing challenges like misclassification, delayed synchronization and evaluated the SyncNet achieved the accuracy of 95% on lip synchronization with labelled dataset.

A brain tumor is an anomalous blowup of cells if not dually detected. A brain excrescence must be found as soon as possible for therapeutic use and survival chances. There are numerous different types of brain tumors with different features, sizes, and treatment choices. Tumor detection manually is difficult, laborious, and prone to mistakes. In this research, deep features are extracted and passed to an ensemble system of AlexNet, EfficientNet, ResNet50, InceptionV3, and VGG16 from which a score vector is acquired from Softmax for demarcation between glioma, meningioma, pituitary, and no tumor. The issues in the published studies have already been estimated on a standard dataset such as Kaggle, 2020-BRATS. The model has detection scores of more than 96%, which shows that the proposed ensemble model outperforms existing works.

Gesture Recognition which is an inevitable part of human computer interaction is an ever evolving active research area. This has been in use under varied applications like smart home systems, sign language recognition, augment reality and in device controls. The acquisition of hand gestures are usually through optical sensors, in this work a radar based hand gesture data set is used for classifying the gestures. Micro doppler signatures were used as the input to the model. Radar dataset has fewer data samples compared to optical based data sets. The proposed work uses separable convolutional neural networks model which does depth wise convolution followed by point wise convolution to reduce overfitting effect of the training data. The proposed model was built in such a way that the model is capable of classifying any unseen data without exactly mimicking the training samples. The proposed model has achieved 94.56% as the testing accuracy which is certainly better than the previous work on this Dop Net data set. Moreover the model has also minimized the computational hours of the model using separable convolutions.

In this paper, we propose a hybrid relay-reflecting intelligent surface (HR-RIS)-assisted cell-free (CF) massive multiple-input multiple-output (mMIMO) network to achieve consistent spectral efficiency (SE). The HR-RIS technique manipulates the propagation environment by reflecting and enhancing radio signals in desired locations, providing a symbiotic integration with CF mMIMO for future wireless communication systems. We first model uplink and downlink channels, obtaining minimum-mean-square-error estimations for efficient transmission paths. We then analyze the SE performance of the proposed system. To enhance sum spectral efficiency in downlink multi-antenna, multi-user, and millimeter-wave massive MIMO networks, we introduce a low-complexity hybrid precoding approach. The optimal analog equalizer is determined by converting the analog precoding matrix dimensions into square matrices and selecting a few discrete Fourier transforms to maximize the amplitude of corresponding wideband channel matrices. We employ the equal gain transmission technique to combine channel gains efficiently and ensure spectral efficiency. To mitigate inter-user interference, we propose an enhanced block diagonalization method for designing the digital precoder and combiner. Our study demonstrates that the proposed HR-RIS-assisted CF mMIMO system offers significant improvements in SE performance, paving the way for advanced wireless communication systems.

Swarm robots, which are inspired from the way insects behave collectively in order to achieve a common goal, have become a major part of research with applications involving search and rescue, area exploration, surveillance etc. In this paper, we present a swarm of robots that do not require individual extrinsic sensors to sense the environment but instead use a single central camera to locate and map the swarm. The robots can be easily built using readily available components with the main chassis being 3D printed, making the system low-cost, low-maintenance, and easy to replicate. We describe Zutu's hardware and software architecture, the algorithms to map the robots to the real world, and some experiments conducted using four of our robots. Eventually, we conclude the possible applications of our system in research, education, and industries.

Human Activity Recognition based research has got intensified based on the evolving demand of smart systems. There has been already a lot of wearables, digital smart sensors deployed to classify various activities. Radar sensor-based Activity recognition has been an active research area during recent times. In order to classify the radar micro doppler signature images we have proposed a approach using Convolutional Neural Network-Long Short Term Memory (CNN-LSTM). Convolutional Layer is used to update the filter values to learn the features of the radar images. LSTM Layer enhances the temporal information besides the features obtained through Convolutional Neural Network. We have used a dataset published by University of Glasgow that captures six activities for 56 subjects under different ages, which is a first of its kind dataset unlike the signals captured under controlled lab environment. Our Model has achieved 96.8% for the training data and 93.5% for the testing data. The proposed work has outperformed the existing traditional deep learning Architectures.

Since the adoption of the internet as a medium of communication of information, fake or false information or news has always been a major issue. Incidents of false information have always increased at times of crisis on national or international scales. The world witnessed a global pandemic from the Coronavirus, causing a complete disruption in the functioning of society. News of bogus cures, home remedies, and medicines started to make their way around the world. The number of incidents of such false news only increased as the pandemic worsened and more people were falling sick and dying. In times of desperation, people can easily be persuaded to try unverified and possibly dangerous medicines or cures, that can cost them their money as well as health. In this paper, natural language processing is used to first identify and differentiate text that has information regarding Covid 19 from the text that does not contain information regarding Covid 19. Word frequency scores like TF and IDF scores are then calculated. The intent of the text is then analyzed by observing the mannerisms detected in false news. With this analysis, the potential of the text to be false or fake is then determined. This research intends to explore the linguistics of false news and to get one step ahead in identifying fake news. The same methodology can be used to analyze data related to other specific topics.

Radar sensing technology that uses Frequency Modulated Continuous Wave Radar(FMCW) has been a promising solution for Human Activity Classification in recent years. Poor monitoring has taken a lot of lives in the recent COVID-19 pandemic which has emphasized the need for better monitoring systems. In this work, we have explored various color spaces and used LAB color space based micro doppler signatures as the Deep Learning model input. We have proposed a novel widened convolutional neural network architecture with parallel input layers for better feature extraction. This helps in yielding classification among 6 Activities of data captured under realistic environments, unlike other radar data sets captured only in lab environments. We have obtained good recognition rates with this architecture that uses LAB based Color space images as input.

Human activity recognition has become an obligatory necessity in day to day life and possible solutions can be provided with the technological advancement of sensing field. Radar based sensing with its unbeatable unique features has been a promising solution for identifying and distinguishing human activities in recent years. The ascent of loss of life among elderly people in care homes during COVID-19 is mainly due to poor monitoring services, that was not able to track their daily life activities. This has even more emphasized the need for savvy activity monitoring and tracking system. In this work, we have used a dataset that has captured six daily life activities of people from different locations during different times under realistic environments, unlike an regular controlled data collection environment. We have proposed a novel tower based convolutional neural network architecture that has employed parallel input layers with individual color channel images sent as inputs to the model. We have concatenated all the unique signature features from each channel to have better and robust feature representation to the model. We have analyzed the proposed model with different color spaces like RGB, LAB, HSV as inputs and found that our chosen input type performs better with the proposed model with significant test accuracy results. We have also compared our proposed model with other existing state of art architectures for radar based human activity recognition.

Deep learning is a field in artificial intelligence that works well in computer vision, natural language processing and audio recognition. Deep neural network architectures has number of layers to conceive the features well, by itself. The hyperparameter tuning plays a major role in every dataset which has major effect in the performance of the training model. Due to the large dimensionality of data it is impossible to tune the parameters by human expertise. In this paper, we have used the CIFAR-10 Dataset and applied the Bayesian hyperparameter optimization algorithm to enhance the performance of the model. Bayesian optimization can be used for any noisy black box function for hyperparameter tuning. In this work Bayesian optimization clearly obtains optimized values for all hyperparameters which saves time and improves performance. The results also show that the error has been reduced in graphical processing unit than in CPU by 6.2% in the validation. Achieving global optimization in the trained model helps transfer learning across domains as well.

The primary objective of this paper is to develop a methodology for brain tumor segmentation. Nowadays, brain tumor recognition and fragmentation is one among the pivotal procedure in surgical and medication planning arrangements. It is difficult to segment the tumor area from MRI images due to inaccessibility of edge and appropriately visible boundaries. In this paper, a combination of Artificial Neural Network and Fuzzy K-means algorithm has been presented to segment the tumor locale. It contains four phases, (1) Noise evacuation (2) Attribute extraction and selection (3) Classification and (4) Segmentation. Initially, the procured image is denoised utilizing wiener filter, and then the significant GLCM attributes are extricated from the images. Then Deep Learning based classification has been performed to classify the abnormal images from the normal images. Finally, it is processed through the Fuzzy K-Means algorithm to segment the tumor region separately. This proposed segmentation approach has been verified on BRATS dataset and produces the accuracy of 94%, sensitivity of 98% specificity of 99%, Jaccard index of 96%. The overall accuracy of this proposed technique has been improved by 8% when compared with K-Nearest Neighbor methodology.

Nowadays wireless sensor networks have wide applications in many fields such as medical, industrial, military, etc. Sensors are having limited processing capabilities and energy is an important constraint in wireless sensor networks as it determines the lifetime of the network.In a large scale wireless sensor networks, the sensor nodes have to collect more data when they are moving towards the sink.So the sensor nodes energy, nearer to the sink may get drained off more quickly and an alternate path has to be chosen which increases the delay in the network.Compressive data gathering is one of the techniques which reduce the data size, balance the energy in large scale wireless sensor networks.The existing technique takes more computations and increase in complexity on compressing the data.So a modified compressive data gathering protocol is designed which uses a lossless compression algorithm called modified Lempel Ziv Welch compression algorithm which is a simple and fast method for compressing the data.A tree is constructed and parent-children nodes are assigned and child node carries the compressed data and intermediate parent node aggregates and compresses it until it reaches the sink.The original data is reconstructed at the sink using Modified Lempel Ziv Welch decompression algorithm Index Terms -Wireless sensor network, data aggregation, MLZW compression and decompression.I. INTRODUCTION Wireless sensor networks consist of numerous autonomous sensors deployed in an environment to monitor physical or environmental conditions such as temperature, humidity etc.They transfer the data to a coordinating node for processing which acts as gateway to pass the data to sink.Each sensor node may consist sensing unit, processing, memory and microcontroller.The size of sensor nodes may vary from grain size to a shoe-box size depending upon the application in which it is used.Similarly, cost may vary from few to hundreds of dollars.The architecture and components of a wireless sensor node is shown in Figure : 1.1.Size and cost constraints of sensor results in resource constraints such as energy consumption, memory, bandwidth and speed.A wireless sensor network is an emerging technology nowadays, which plays an important role in creating a smart environment.It has huge applications in many fields such as environmental monitoring, natural hazard detection, military applications, health monitoring, forest monitoring, precision agriculture etc.