Anubha Parashar | Manipal University Jaipur (original) (raw)

Uploads

Papers by Anubha Parashar

Image and Vision Computing

Journal of Innovation in Computer Science and Engineering, 2020

Cognitive Computing for Human-Robot Interaction, 2021

Abstract Autonomous driving has passed the point of being called the biggest step, as the smart c... more Abstract Autonomous driving has passed the point of being called the biggest step, as the smart car revolution is already taking shape around the world. Self-driving cars are relevant if not prevalent and the biggest obstacles to reach the mass adoption are customer acceptance, cost, infrastructure, and the reliance on several onerous algorithms that include perception, lane marking detection, path planning, and variation in pathways. This study tackled the mentioned problems with a straightforward and cost-effective solution, using end-to-end learning and replacing the numerous sensors with a camera and commandeering just the forward, backward, left, and right controls. In this research, authors have used the most popular method of deep learning that is convolutional neural network to train the collected data on the VGG16 model. Later these have optimized directly by the proposed system with cropping each unnecessary image and mapping pixels from a single front-facing camera to direct steering instructions. It has been observed from the experimental work that the proposed model has given a better result than the existing work, that is, increase in the accuracy from 88% (Udacity training dataset) to 98% (proposed). This model is suitable for industrial use and robust in real time scenarios, therefore can be applied in modern industrialized systems.

Smart Systems and IoT: Innovations in Computing, 2019

E-nose is a self-controlled smoke detector developed using Arduino microcontroller and sensors. I... more E-nose is a self-controlled smoke detector developed using Arduino microcontroller and sensors. It is capable of sensing the smell in the ambiance via gas sensors and programmed to alert the smokers giving red signal with buzzer. Smoke detection algorithm facilitates the intelligent warning to the cigarette smokers. Nowadays, smokers’ count is increasing with a high-speed raising health and moral issues to the people especially in case of passive smoking at public places. Today “No Smoking” rules are only written, but to make these rules being followed this work would be a milestone. To implement e-Nose, a cigarette smoke detector with red/green LED and buzzer using Arduino uno board R3 is developed. MQ-2 gas sensor is used to detect cigarette smoke and combination of buzzer and LEDs is used to warn smokers. This work demonstrates the development of e-Nose and evaluation of the performance of developed product. In nutshell, this e-Nose which is cigarette smoke detector is designed to achieve high degree of awareness among public and to reduce the smokers’ count especially at public places. We detect the smoke through e-Nose and upload on the cloud. The cloud analysis is done using various machine learning techniques to categorize the smoke with pollution data. And result shows smoking percentage.

Intelligent Data Communication Technologies and Internet of Things, 2022

Security and Trust Issues in Internet of Things, 2020

These days’ security plays the greatest role in order to provide safe platforms and surveillance ... more These days’ security plays the greatest role in order to provide safe platforms and surveillance becomes the essential need to provide accurate results in case of security breach. Gait recognition is a biometric technique that does not need human intervention. Through this technique, human can be uniquely identified. Gait is defined as the way human walks (human locomotion) and this can be used as biometric identity because the manner in which every person walks can uniquely categorize person. But there are many challenges like variation in viewpoints, clothing variations, carrying conditions and so on. A novel approach using deep learning is proposed to address this challenge. To address these limits, we give an idea of having multi-view gait-based recognition system, to provide robust system that is capable of handling one camera and subject walking on different angles from 0° to 180° of view. To achieve the results 3D CNN-based model is used in order to obtain spatio-temporal fea...

The Classification of Humanoid locomotion is a troublesome exercise because of nonlinearity assoc... more The Classification of Humanoid locomotion is a troublesome exercise because of nonlinearity associate with gait. The high dimension feature vector requires a high computational cost. The classification using the different machine learning technique leads for over fitting and under fitting. To select the correct feature is also the difficult task. The hand craft feature selection machine learning techniques performed poor. We have used the deep learning technique to get the trained feature and then classification we have used deep belief network-based deep learning. Classification is utilized to see Gait pattern of different person and any upcoming disease can be detected earlier. So in this paper we first selected the feature and identify the principle feature then we classify gait data and use different machine learning technique (ANN, SVM, KNN, and Classifier fusion) and performance comparison is shown. Experimental result on real time datasets propose method is better than previo...

Smart Systems and IoT: Innovations in Computing, 2019

In the age of digitalization, Internet-based applications are gaining popularity at an exponentia... more In the age of digitalization, Internet-based applications are gaining popularity at an exponential rate. Today, everyone wants to make their lives easier and devices smarter. In the age of automation, most of the devices we interact with on a day-to-day basis, for example, air conditioners, refrigerators, etc. are made increasingly intelligent to simplify our lives and make it comfortable. Using the principles of IoT and AI, we can create home automation devices such as automatic security devices and e-meters that make our homes smarter and more secure. Keeping a track of how much electricity is consumed per household becomes imperative seeing the rate at which global warming is increasing. Gone are the days where users had to go to meter reading room and take down readings. By employing IoT concepts, we can simplify this tedious process and record the reading over cloud for easy accessibility. The major advantage of digitalizing the process is that the user has the facility to view...

Autonomous Driving and Advanced Driver-Assistance Systems (ADAS), 2021

Autonomous Driving and Advanced Driver-Assistance Systems (ADAS), 2021

2019 2nd International Conference on Intelligent Communication and Computational Techniques (ICCT), 2019

The weather can have great impact on lives. Weather changes can influence wide range of human act... more The weather can have great impact on lives. Weather changes can influence wide range of human activities and affect agriculture and transportation. The main aim of this paper is to monitor and report weather conditions so that one is informed beforehand and necessary actions can be taken to reduce the damage by any calamity by forecasting it. Here we are using various sensors in order to collect the data and previous data is used in order to train the system and with current data collection we do the prediction. We will be analyzing temperature, pollutants, humidity and pressure and will predict the weather. Existing models are expensive in contrast to ours and hence it will make monitoring local area feasible as it will be cheaper.

International Journal of Information Technology and Computer Science, 2018

Millions of companies expend billions of dollars on trillions of software for the development and... more Millions of companies expend billions of dollars on trillions of software for the development and maintenance. Still many projects result in failure causing heavy financial loss. Major reason is the inefficient effort estimation techniques which are not so suitable for the current development methods. The continuous change in the software development technology makes effort estimation more challenging. Till date, no estimation method has been found full-proof to accurately pre-compute the time, money, effort (man-hours) and other resources required to successfully complete the project resulting either overestimated budget or underestimated budget. Here a machine learning COCOMO is proposed which is a novel non-algorithmic approach to effort estimation. This estimation technique performs well within their pre-specified domains and beyond so. As development methods have undergone revolutionaries but estimation techniques are not so modified to cope up with the modern development skills, so the need of training the models to work with updated development methods is being satiated just by finding out the patterns and associations among the domain specific data sets via neural networks along with carriage of desired COCOMO features. This paper estimates the effort by training proposed neural network using already published data-set and later on, the testing is done. The validation clearly shows that the performance of algorithmic method is improved by the proposed machine learning method.

International Journal of Wireless and Microwave Technologies, 2017

Communications in Computer and Information Science, 2016

Clustering is done in order to group the entities which are alike in one group so that grouping o... more Clustering is done in order to group the entities which are alike in one group so that grouping of more similar objects can be done. The objects placed in one group are known as clusters. In this paper we are using clustering in order to identify the human locomotion and categories the dataset making clusters. We are using two clustering techniques i.e. SOM and K-mean. So we first selected the feature and identify the principle feature then we cluster gait data and use different machine learning technique (K-mean and SOM) and performance comparison is shown. Experimental result on real time datasets propose method is better than previous method as far as humanoid locomotion classification is concerned.

International Journal of Interactive Multimedia and Artificial Intelligence, 2016

Push recovery is prime ability that is essential to be incorporated in the process of developing ... more Push recovery is prime ability that is essential to be incorporated in the process of developing a robust humanoid robot to support bipedalism. In real environment it is very essential for humanoid robot to maintain balance. In this paper we are generating a control system and push recovery controller for humanoid robot walking. We apply different kind of pushes to humanoid robot and the algorithm that can bring a change in the walking stage to sustain walking. The simulation is done in 3D environment using Webots. This paper describes techniques for feature selection to foreshow push recovery for hip, ankle and knee joint. We train the system by K-Mean algorithm and testing is done on crouch data and tested results are reported. Random push data of humanoid robot is collected and classified to see whether push lie in safer region and then tested on given proposed system.

Image and Vision Computing

Journal of Innovation in Computer Science and Engineering, 2020

Cognitive Computing for Human-Robot Interaction, 2021

Abstract Autonomous driving has passed the point of being called the biggest step, as the smart c... more Abstract Autonomous driving has passed the point of being called the biggest step, as the smart car revolution is already taking shape around the world. Self-driving cars are relevant if not prevalent and the biggest obstacles to reach the mass adoption are customer acceptance, cost, infrastructure, and the reliance on several onerous algorithms that include perception, lane marking detection, path planning, and variation in pathways. This study tackled the mentioned problems with a straightforward and cost-effective solution, using end-to-end learning and replacing the numerous sensors with a camera and commandeering just the forward, backward, left, and right controls. In this research, authors have used the most popular method of deep learning that is convolutional neural network to train the collected data on the VGG16 model. Later these have optimized directly by the proposed system with cropping each unnecessary image and mapping pixels from a single front-facing camera to direct steering instructions. It has been observed from the experimental work that the proposed model has given a better result than the existing work, that is, increase in the accuracy from 88% (Udacity training dataset) to 98% (proposed). This model is suitable for industrial use and robust in real time scenarios, therefore can be applied in modern industrialized systems.

Smart Systems and IoT: Innovations in Computing, 2019

E-nose is a self-controlled smoke detector developed using Arduino microcontroller and sensors. I... more E-nose is a self-controlled smoke detector developed using Arduino microcontroller and sensors. It is capable of sensing the smell in the ambiance via gas sensors and programmed to alert the smokers giving red signal with buzzer. Smoke detection algorithm facilitates the intelligent warning to the cigarette smokers. Nowadays, smokers’ count is increasing with a high-speed raising health and moral issues to the people especially in case of passive smoking at public places. Today “No Smoking” rules are only written, but to make these rules being followed this work would be a milestone. To implement e-Nose, a cigarette smoke detector with red/green LED and buzzer using Arduino uno board R3 is developed. MQ-2 gas sensor is used to detect cigarette smoke and combination of buzzer and LEDs is used to warn smokers. This work demonstrates the development of e-Nose and evaluation of the performance of developed product. In nutshell, this e-Nose which is cigarette smoke detector is designed to achieve high degree of awareness among public and to reduce the smokers’ count especially at public places. We detect the smoke through e-Nose and upload on the cloud. The cloud analysis is done using various machine learning techniques to categorize the smoke with pollution data. And result shows smoking percentage.

Intelligent Data Communication Technologies and Internet of Things, 2022

Security and Trust Issues in Internet of Things, 2020

These days’ security plays the greatest role in order to provide safe platforms and surveillance ... more These days’ security plays the greatest role in order to provide safe platforms and surveillance becomes the essential need to provide accurate results in case of security breach. Gait recognition is a biometric technique that does not need human intervention. Through this technique, human can be uniquely identified. Gait is defined as the way human walks (human locomotion) and this can be used as biometric identity because the manner in which every person walks can uniquely categorize person. But there are many challenges like variation in viewpoints, clothing variations, carrying conditions and so on. A novel approach using deep learning is proposed to address this challenge. To address these limits, we give an idea of having multi-view gait-based recognition system, to provide robust system that is capable of handling one camera and subject walking on different angles from 0° to 180° of view. To achieve the results 3D CNN-based model is used in order to obtain spatio-temporal fea...

The Classification of Humanoid locomotion is a troublesome exercise because of nonlinearity assoc... more The Classification of Humanoid locomotion is a troublesome exercise because of nonlinearity associate with gait. The high dimension feature vector requires a high computational cost. The classification using the different machine learning technique leads for over fitting and under fitting. To select the correct feature is also the difficult task. The hand craft feature selection machine learning techniques performed poor. We have used the deep learning technique to get the trained feature and then classification we have used deep belief network-based deep learning. Classification is utilized to see Gait pattern of different person and any upcoming disease can be detected earlier. So in this paper we first selected the feature and identify the principle feature then we classify gait data and use different machine learning technique (ANN, SVM, KNN, and Classifier fusion) and performance comparison is shown. Experimental result on real time datasets propose method is better than previo...

Smart Systems and IoT: Innovations in Computing, 2019

In the age of digitalization, Internet-based applications are gaining popularity at an exponentia... more In the age of digitalization, Internet-based applications are gaining popularity at an exponential rate. Today, everyone wants to make their lives easier and devices smarter. In the age of automation, most of the devices we interact with on a day-to-day basis, for example, air conditioners, refrigerators, etc. are made increasingly intelligent to simplify our lives and make it comfortable. Using the principles of IoT and AI, we can create home automation devices such as automatic security devices and e-meters that make our homes smarter and more secure. Keeping a track of how much electricity is consumed per household becomes imperative seeing the rate at which global warming is increasing. Gone are the days where users had to go to meter reading room and take down readings. By employing IoT concepts, we can simplify this tedious process and record the reading over cloud for easy accessibility. The major advantage of digitalizing the process is that the user has the facility to view...

Autonomous Driving and Advanced Driver-Assistance Systems (ADAS), 2021

Autonomous Driving and Advanced Driver-Assistance Systems (ADAS), 2021

2019 2nd International Conference on Intelligent Communication and Computational Techniques (ICCT), 2019

The weather can have great impact on lives. Weather changes can influence wide range of human act... more The weather can have great impact on lives. Weather changes can influence wide range of human activities and affect agriculture and transportation. The main aim of this paper is to monitor and report weather conditions so that one is informed beforehand and necessary actions can be taken to reduce the damage by any calamity by forecasting it. Here we are using various sensors in order to collect the data and previous data is used in order to train the system and with current data collection we do the prediction. We will be analyzing temperature, pollutants, humidity and pressure and will predict the weather. Existing models are expensive in contrast to ours and hence it will make monitoring local area feasible as it will be cheaper.

International Journal of Information Technology and Computer Science, 2018

Millions of companies expend billions of dollars on trillions of software for the development and... more Millions of companies expend billions of dollars on trillions of software for the development and maintenance. Still many projects result in failure causing heavy financial loss. Major reason is the inefficient effort estimation techniques which are not so suitable for the current development methods. The continuous change in the software development technology makes effort estimation more challenging. Till date, no estimation method has been found full-proof to accurately pre-compute the time, money, effort (man-hours) and other resources required to successfully complete the project resulting either overestimated budget or underestimated budget. Here a machine learning COCOMO is proposed which is a novel non-algorithmic approach to effort estimation. This estimation technique performs well within their pre-specified domains and beyond so. As development methods have undergone revolutionaries but estimation techniques are not so modified to cope up with the modern development skills, so the need of training the models to work with updated development methods is being satiated just by finding out the patterns and associations among the domain specific data sets via neural networks along with carriage of desired COCOMO features. This paper estimates the effort by training proposed neural network using already published data-set and later on, the testing is done. The validation clearly shows that the performance of algorithmic method is improved by the proposed machine learning method.

International Journal of Wireless and Microwave Technologies, 2017

Communications in Computer and Information Science, 2016

Clustering is done in order to group the entities which are alike in one group so that grouping o... more Clustering is done in order to group the entities which are alike in one group so that grouping of more similar objects can be done. The objects placed in one group are known as clusters. In this paper we are using clustering in order to identify the human locomotion and categories the dataset making clusters. We are using two clustering techniques i.e. SOM and K-mean. So we first selected the feature and identify the principle feature then we cluster gait data and use different machine learning technique (K-mean and SOM) and performance comparison is shown. Experimental result on real time datasets propose method is better than previous method as far as humanoid locomotion classification is concerned.

International Journal of Interactive Multimedia and Artificial Intelligence, 2016

Push recovery is prime ability that is essential to be incorporated in the process of developing ... more Push recovery is prime ability that is essential to be incorporated in the process of developing a robust humanoid robot to support bipedalism. In real environment it is very essential for humanoid robot to maintain balance. In this paper we are generating a control system and push recovery controller for humanoid robot walking. We apply different kind of pushes to humanoid robot and the algorithm that can bring a change in the walking stage to sustain walking. The simulation is done in 3D environment using Webots. This paper describes techniques for feature selection to foreshow push recovery for hip, ankle and knee joint. We train the system by K-Mean algorithm and testing is done on crouch data and tested results are reported. Random push data of humanoid robot is collected and classified to see whether push lie in safer region and then tested on given proposed system.