The Fog-to-Cloud (F2C) paradigm is emerging to both provide higher functional efficiency for latency-sensitive services and also help modern computing systems to be more intelligent. As it is still in its infancy, the biggest challenge for this domain is to build a proper resource allocation technique as part of an efficient resource management module. The diversified and distributed nature of that paradigm creates some additional hurdles for choosing the appropriate resources for executing some tasks. Significantly, efficient resource consumption estimation and performance forecasting are core issues in the design and development of a proper and smart resource management mechanism for F2C systems. Considering this fact, in this paper, we aim at designing an architectural framework for a prediction-based resource management mechanism for F2C systems. The performance prediction is based on supervised machine learning technology. The proposal has been evaluated and validated by predicting the performance and resources usage of F2C resources through several tests. Primarily, we have run an image recognition application on different F2C resources and collected performance-related information and resource consumption information. Then, by adopting the multivariate regression methodology, we perform some standard machine learning techniques to predict the performance and estimate the resource consumption of the F2C resources. Finally, to justify the effectiveness of our proposal, we calculated the value of a cost function between estimated values and the real measured values.
{"title":"An Architectural Schema for Performance Prediction using Machine Learning in the Fog-to-Cloud Paradigm","authors":"Souvik Sengupta, Jordi García, X. Masip-Bruin, Andrés Prieto-González","doi":"10.1109/UEMCON47517.2019.8992939","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992939","url":null,"abstract":"The Fog-to-Cloud (F2C) paradigm is emerging to both provide higher functional efficiency for latency-sensitive services and also help modern computing systems to be more intelligent. As it is still in its infancy, the biggest challenge for this domain is to build a proper resource allocation technique as part of an efficient resource management module. The diversified and distributed nature of that paradigm creates some additional hurdles for choosing the appropriate resources for executing some tasks. Significantly, efficient resource consumption estimation and performance forecasting are core issues in the design and development of a proper and smart resource management mechanism for F2C systems. Considering this fact, in this paper, we aim at designing an architectural framework for a prediction-based resource management mechanism for F2C systems. The performance prediction is based on supervised machine learning technology. The proposal has been evaluated and validated by predicting the performance and resources usage of F2C resources through several tests. Primarily, we have run an image recognition application on different F2C resources and collected performance-related information and resource consumption information. Then, by adopting the multivariate regression methodology, we perform some standard machine learning techniques to predict the performance and estimate the resource consumption of the F2C resources. Finally, to justify the effectiveness of our proposal, we calculated the value of a cost function between estimated values and the real measured values.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"608 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132858662","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8992992
Shajina Anand, G. Raja, Aishwarya Ganapathisubramaniyan
Due to enormous growth of devices and its interconnected communication, there are lot of possibilities for attacks and vulnerabilities in the network. From recent studies, we come to know about various security related techniques and their usage. A lot of cryptographic algorithms are used efficiently and effectively in plenty of applications. Thus it is notable to classify the attacks as two types Active and Passive Attack. Unauthorized access to the information implies passive attack. For example, in the communication channel doing operations like intercepting and eavesdropping are considered as passive attack. If the attacker tries to access the information and modify the information in an unauthorized manner, then it is considered as active attack. Man in the Middle attack is the best example of active attack. Here the attacker sends its public key as the hosts public and tries to hack the message. To overcome the issues in tradional authentication model, we proposed an Advanced Elliptical Curve Cryptographic (AECC) Algorithm which uses a code bit with a random function in both the sender and receiver side for encoding and decoding functions. This code bit is used to identify whether the public key belongs to attacker or the sender. By identifying the sender, the receiver can communicate with it without any restrictions. The results analysis section shows that our proposed AECC algorithm enhances security for user defined input messages like text, image, video and also for binary data.
{"title":"AECC: An Enhanced Public Key Cryptosystem for User Defined Messages","authors":"Shajina Anand, G. Raja, Aishwarya Ganapathisubramaniyan","doi":"10.1109/UEMCON47517.2019.8992992","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992992","url":null,"abstract":"Due to enormous growth of devices and its interconnected communication, there are lot of possibilities for attacks and vulnerabilities in the network. From recent studies, we come to know about various security related techniques and their usage. A lot of cryptographic algorithms are used efficiently and effectively in plenty of applications. Thus it is notable to classify the attacks as two types Active and Passive Attack. Unauthorized access to the information implies passive attack. For example, in the communication channel doing operations like intercepting and eavesdropping are considered as passive attack. If the attacker tries to access the information and modify the information in an unauthorized manner, then it is considered as active attack. Man in the Middle attack is the best example of active attack. Here the attacker sends its public key as the hosts public and tries to hack the message. To overcome the issues in tradional authentication model, we proposed an Advanced Elliptical Curve Cryptographic (AECC) Algorithm which uses a code bit with a random function in both the sender and receiver side for encoding and decoding functions. This code bit is used to identify whether the public key belongs to attacker or the sender. By identifying the sender, the receiver can communicate with it without any restrictions. The results analysis section shows that our proposed AECC algorithm enhances security for user defined input messages like text, image, video and also for binary data.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133588017","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8992967
Brian Luu, Bradley Hansberger, T. Tothong, K. George
This paper presents evidence for the possibility of a neuroadaptive system, based on electroencephalography (EEG) readings from the motor cortex region, to predict an individual's actions before the onset of motion. Testing for the neuroadaptive system utilized a G.nautilus headset, MATLAB with the EEGLAB toolbox, and a computer with the Processing IDE. Code from the Processing IDE provides an image slideshow which displays faces of various individuals so that the subject presses a keyboard key on a certain image. Three subjects were tested for 60 trials each, 30 trials where the keyboard key was pressed, and 30 trials where they were not, to gather enough data to train and test a classifier by using a machine learning algorithm. Machine learning assessed classification accuracy initially using 10 training trials and increased the training set by 10 trials each time to reassess accuracy until a total of 40 training trials were used. A set of 20 trials were used to assess accuracy without and with machine learning. Additionally, theoretical accuracy was computed by removing unfeasible machine learning features to assess the potential accuracy in a real-time system. The results provided an average accuracy of 52% without machine learning and an average accuracy ranging from 91.66% to 96.66% using the K-Nearest Neighbors(KNN) algorithm. The average theoretical accuracy ranged from to be 60% to 68.33%.
{"title":"Conceptual Neuroadaptive Brain-Computer Interface utilizing Event-related Desynchronization","authors":"Brian Luu, Bradley Hansberger, T. Tothong, K. George","doi":"10.1109/UEMCON47517.2019.8992967","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992967","url":null,"abstract":"This paper presents evidence for the possibility of a neuroadaptive system, based on electroencephalography (EEG) readings from the motor cortex region, to predict an individual's actions before the onset of motion. Testing for the neuroadaptive system utilized a G.nautilus headset, MATLAB with the EEGLAB toolbox, and a computer with the Processing IDE. Code from the Processing IDE provides an image slideshow which displays faces of various individuals so that the subject presses a keyboard key on a certain image. Three subjects were tested for 60 trials each, 30 trials where the keyboard key was pressed, and 30 trials where they were not, to gather enough data to train and test a classifier by using a machine learning algorithm. Machine learning assessed classification accuracy initially using 10 training trials and increased the training set by 10 trials each time to reassess accuracy until a total of 40 training trials were used. A set of 20 trials were used to assess accuracy without and with machine learning. Additionally, theoretical accuracy was computed by removing unfeasible machine learning features to assess the potential accuracy in a real-time system. The results provided an average accuracy of 52% without machine learning and an average accuracy ranging from 91.66% to 96.66% using the K-Nearest Neighbors(KNN) algorithm. The average theoretical accuracy ranged from to be 60% to 68.33%.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"94 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121833078","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8993032
Mahssa Abdolahi, Hao Jiang, B. Kaminska
In this work, robust readout of the data (232 English characters) stored in high-security structural colour QR codes, was achieved by using multiple image processing techniques, specifically, histogram equalization and decorrelation stretching. The decoded structural colour QR codes are generic diffractive RGB-pixelated periodic nanocones selectively activated by laser exposure to obtain the particular design of interest. The samples were imaged according to the criteria determined by the diffraction grating equation for the lighting and viewing angles given the red, green, and blue periodicities of the grating. However, illumination variations all through the samples, cross-module and cross-channel interference effects result in acquiring images with dissimilar lighting conditions which cannot be directly retrieved by the decoding script and need significant preprocessing. According to the intensity plots, even if the intensity values are very close (above ~200) at some typical regions of the images with different lighting conditions, their inconsistencies (below ~100) at the pixels of one representative region may lead to the requirement for using different methods for recovering the data from all red, green, and blue channels. In many cases, a successful data readout could be achieved by downscaling the images to ~300-pixel dimensions (along with bilinear interpolation resampling), histogram equalization (HE), linear spatial low-pass mean filtering, and gamma function, each used either independently or with other complementary processes. The majority of images, however, could be fully decoded using decorrelation stretching (DS) either as a standalone or combinational process for obtaining a more distinctive colour definition.
{"title":"Robust data retrieval from high-security structural colour QR codes via histogram equalization and decorrelation stretching","authors":"Mahssa Abdolahi, Hao Jiang, B. Kaminska","doi":"10.1109/UEMCON47517.2019.8993032","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8993032","url":null,"abstract":"In this work, robust readout of the data (232 English characters) stored in high-security structural colour QR codes, was achieved by using multiple image processing techniques, specifically, histogram equalization and decorrelation stretching. The decoded structural colour QR codes are generic diffractive RGB-pixelated periodic nanocones selectively activated by laser exposure to obtain the particular design of interest. The samples were imaged according to the criteria determined by the diffraction grating equation for the lighting and viewing angles given the red, green, and blue periodicities of the grating. However, illumination variations all through the samples, cross-module and cross-channel interference effects result in acquiring images with dissimilar lighting conditions which cannot be directly retrieved by the decoding script and need significant preprocessing. According to the intensity plots, even if the intensity values are very close (above ~200) at some typical regions of the images with different lighting conditions, their inconsistencies (below ~100) at the pixels of one representative region may lead to the requirement for using different methods for recovering the data from all red, green, and blue channels. In many cases, a successful data readout could be achieved by downscaling the images to ~300-pixel dimensions (along with bilinear interpolation resampling), histogram equalization (HE), linear spatial low-pass mean filtering, and gamma function, each used either independently or with other complementary processes. The majority of images, however, could be fully decoded using decorrelation stretching (DS) either as a standalone or combinational process for obtaining a more distinctive colour definition.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116927006","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8992932
Jackie Leung, Min Chen
Convolutional neural networks (CNNs) have gained global recognition in advancing the field of artificial intelligence and have had great successes in a wide array of applications including computer vision, speech and natural language processing. However, due to the rise of big data and increased complexity of tasks, the efficiency of training CNNs have been severely impacted. To achieve state-of-art results, CNNs require tens to hundreds of millions of parameters that need to be fine-tuned, resulting in extensive training time and high computational cost. To overcome these obstacles, this work takes advantage of distributed frameworks and cloud computing to develop a parallel CNN algorithm. Close examination of the implementation of MapReduce based CNNs as well as how the proposed algorithm accelerates learning are discussed and demonstrated through experiments. Results reveal high accuracy in classification and improvements in speedup, scaleup and sizeup compared to the standard algorithm.
{"title":"Image Recognition with MapReduce Based Convolutional Neural Networks","authors":"Jackie Leung, Min Chen","doi":"10.1109/UEMCON47517.2019.8992932","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992932","url":null,"abstract":"Convolutional neural networks (CNNs) have gained global recognition in advancing the field of artificial intelligence and have had great successes in a wide array of applications including computer vision, speech and natural language processing. However, due to the rise of big data and increased complexity of tasks, the efficiency of training CNNs have been severely impacted. To achieve state-of-art results, CNNs require tens to hundreds of millions of parameters that need to be fine-tuned, resulting in extensive training time and high computational cost. To overcome these obstacles, this work takes advantage of distributed frameworks and cloud computing to develop a parallel CNN algorithm. Close examination of the implementation of MapReduce based CNNs as well as how the proposed algorithm accelerates learning are discussed and demonstrated through experiments. Results reveal high accuracy in classification and improvements in speedup, scaleup and sizeup compared to the standard algorithm.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114662933","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8992993
Resham Jhangiani, Doina Bein, Abhishek Verma
Fraud has become a major problem in e-commerce and a lot of resources are being invested to recognize and prevent it. Present fraud detection and prevention systems are designed to prevent only a small fraction of fraudulent transactions processed, which still costs billions of dollars in loss. There is an urgent need for better fraud detection and prevention as the online transactions are estimated to increase substantially in the coming year. We propose a data driven model using machine learning algorithms on big data to predict the probability of a transaction being fraudulent or legitimate. The model was trained on historical e-commerce credit card transaction data to predict the probability of any future transaction by the customer being fraudulent. Supervised machine learning algorithms like Random Forest, Support Vector Machine, Gradient Boost and combinations of these are implemented and their performance are compared. While at the same time the problem of class imbalance is taken into consideration and techniques of oversampling and data pre-processing are performed before the model is trained on a classifier.
{"title":"Machine Learning Pipeline for Fraud Detection and Prevention in E-Commerce Transactions","authors":"Resham Jhangiani, Doina Bein, Abhishek Verma","doi":"10.1109/UEMCON47517.2019.8992993","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992993","url":null,"abstract":"Fraud has become a major problem in e-commerce and a lot of resources are being invested to recognize and prevent it. Present fraud detection and prevention systems are designed to prevent only a small fraction of fraudulent transactions processed, which still costs billions of dollars in loss. There is an urgent need for better fraud detection and prevention as the online transactions are estimated to increase substantially in the coming year. We propose a data driven model using machine learning algorithms on big data to predict the probability of a transaction being fraudulent or legitimate. The model was trained on historical e-commerce credit card transaction data to predict the probability of any future transaction by the customer being fraudulent. Supervised machine learning algorithms like Random Forest, Support Vector Machine, Gradient Boost and combinations of these are implemented and their performance are compared. While at the same time the problem of class imbalance is taken into consideration and techniques of oversampling and data pre-processing are performed before the model is trained on a classifier.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130741342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8993010
M. Royer, S. Chawathe
We present an approach for scientific data management systems to apply certificates to scientific objects, which are typically unformatted datasets, to facilitate analysis by climate scientists. Typically, for a program to process data, the program requires cleansed data in a form that supports automatic manipulation. Most systems require that data must adhere to a specific format to achieve that goal. The technique described in this paper takes the opposite approach; instead, any dataset may be imported and manipulated in the system. But upon initial import, however, only a subset of system functions may work with any given dataset. As the data is refined and transformed by system functions, more functions may become compatible. Certificates are associated with objects that pass constraint validation within the system to ensure that they conform to function requirements. The attached object constraints represent invariant properties of the object, which may be used by functions in the system as function preconditions. Furthermore, the functions defined in the system may associate certificates with the newly generated results. Certificates related to function results are effectively function postconditions, which in turn are used to associate certificates with the objects generated in the system. Additionally, attached object certificates reflect the refinement of data into a more pristine version. This paper describes the technique for modeling and enforcing the constraints for data scientists that have similar requirements.
{"title":"Unformatted, Certified Scientific Objects","authors":"M. Royer, S. Chawathe","doi":"10.1109/UEMCON47517.2019.8993010","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8993010","url":null,"abstract":"We present an approach for scientific data management systems to apply certificates to scientific objects, which are typically unformatted datasets, to facilitate analysis by climate scientists. Typically, for a program to process data, the program requires cleansed data in a form that supports automatic manipulation. Most systems require that data must adhere to a specific format to achieve that goal. The technique described in this paper takes the opposite approach; instead, any dataset may be imported and manipulated in the system. But upon initial import, however, only a subset of system functions may work with any given dataset. As the data is refined and transformed by system functions, more functions may become compatible. Certificates are associated with objects that pass constraint validation within the system to ensure that they conform to function requirements. The attached object constraints represent invariant properties of the object, which may be used by functions in the system as function preconditions. Furthermore, the functions defined in the system may associate certificates with the newly generated results. Certificates related to function results are effectively function postconditions, which in turn are used to associate certificates with the objects generated in the system. Additionally, attached object certificates reflect the refinement of data into a more pristine version. This paper describes the technique for modeling and enforcing the constraints for data scientists that have similar requirements.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"80 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130381139","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8993079
Mark Caswell, Yao Liang
Communication links in low-power wireless visual sensor networks (WVSNs) are subject to short-term and long-term noise variations. These variations can cause a WVSN to exhibit prolonged or periodic transitional or bursty transmission performance. In this paper, we present our work on how to generate noise traces that simulate real-world transitional and bursty network behavior in TOSSIM. We develop a toolset called BurstyGen for TOSSIM which can facilitate WVSN protocol designers and application developers to better understand WVSN performance under these conditions. BurstyGen allows users to model both short time-scale and long time-scale variations in WVSN noise environments for the simulation and testing of WVSN system algorithms and protocols.
{"title":"Simulation of Transitional and Bursty Wireless Visual Sensor Networks","authors":"Mark Caswell, Yao Liang","doi":"10.1109/UEMCON47517.2019.8993079","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8993079","url":null,"abstract":"Communication links in low-power wireless visual sensor networks (WVSNs) are subject to short-term and long-term noise variations. These variations can cause a WVSN to exhibit prolonged or periodic transitional or bursty transmission performance. In this paper, we present our work on how to generate noise traces that simulate real-world transitional and bursty network behavior in TOSSIM. We develop a toolset called BurstyGen for TOSSIM which can facilitate WVSN protocol designers and application developers to better understand WVSN performance under these conditions. BurstyGen allows users to model both short time-scale and long time-scale variations in WVSN noise environments for the simulation and testing of WVSN system algorithms and protocols.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125791937","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8992929
Heekyung Kim, K. Choi
This paper shows the possibility of the existing low power register transfer level (RTL) techniques can be effective as a low power design scheme for CNN-based object recognition system acceleration in contrast to conventional techniques. Most power efficient design techniques regarding CNN acceleration are focused on the High-level Synthesis (HLS) aspect, such as memory bandwidth optimization, network architecture reconfiguration, data reuse, and batch normalization. However, these attempts have reached the limits of the effectiveness of itself. Using the post-synthesis RTL code generated by field-programmable gate arrays (FPGA) manufacturers, the proposed RTL low power design technique was applied to the original FIFO part for reducing the power consumption during data transformation. We compared the HLS optimized result with the RTL optimized result in the aspect of power consumption. We configured the testbench for the modified FIFO module and analyzed the estimated power dissipation result. These power effectiveness factors, such as a look-up table (LUT), a lookup table RAM (LUTRAM), can reduce the power dissipation by 54%, 49% respectively, even though increased block RAM (BRAM) leads to the elevated power dissipation by 154%. Thus, the total power consumption was able to be decreased by 10%. This paper discusses two factors of FPGA with system-on-chip (FPGA-SoC) design for CNN-based hardware implementation in power consumption aspect, such as RTL architecture, memory design architecture, and the model architecture-based hardware implementation methods. The virtual additional memory can support the high throughput at full speed. Our simulated low power schemes applied to Processing System (PS) and Programmable Logic (PL) architecture effectively reduced the power consumption by 25.9% in the FIFO data transformation. We established that the increased LUT blocks affect the power-efficient rate and reduce the power consumption of the PL design up to 49%.
{"title":"Low Power FPGA-SoC Design Techniques for CNN-based Object Detection Accelerator","authors":"Heekyung Kim, K. Choi","doi":"10.1109/UEMCON47517.2019.8992929","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992929","url":null,"abstract":"This paper shows the possibility of the existing low power register transfer level (RTL) techniques can be effective as a low power design scheme for CNN-based object recognition system acceleration in contrast to conventional techniques. Most power efficient design techniques regarding CNN acceleration are focused on the High-level Synthesis (HLS) aspect, such as memory bandwidth optimization, network architecture reconfiguration, data reuse, and batch normalization. However, these attempts have reached the limits of the effectiveness of itself. Using the post-synthesis RTL code generated by field-programmable gate arrays (FPGA) manufacturers, the proposed RTL low power design technique was applied to the original FIFO part for reducing the power consumption during data transformation. We compared the HLS optimized result with the RTL optimized result in the aspect of power consumption. We configured the testbench for the modified FIFO module and analyzed the estimated power dissipation result. These power effectiveness factors, such as a look-up table (LUT), a lookup table RAM (LUTRAM), can reduce the power dissipation by 54%, 49% respectively, even though increased block RAM (BRAM) leads to the elevated power dissipation by 154%. Thus, the total power consumption was able to be decreased by 10%. This paper discusses two factors of FPGA with system-on-chip (FPGA-SoC) design for CNN-based hardware implementation in power consumption aspect, such as RTL architecture, memory design architecture, and the model architecture-based hardware implementation methods. The virtual additional memory can support the high throughput at full speed. Our simulated low power schemes applied to Processing System (PS) and Programmable Logic (PL) architecture effectively reduced the power consumption by 25.9% in the FIFO data transformation. We established that the increased LUT blocks affect the power-efficient rate and reduce the power consumption of the PL design up to 49%.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"166 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127200032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2019-10-01DOI: 10.1109/UEMCON47517.2019.8992950
Ushashi Chowdhury, Pranjal Chowdhury, Sourav Paul, Anwesha Sen, Partho Protim Sarkar, S. Basak, Abari Bhattacharya
Now-a-days we can see that human life is becoming very fast. Moreover, the city life is getting very busy day- by-day. So in the daily busy schedule it is becoming very difficult for the parents to monitor their children closely. This paper discusses about a smart wearable device like a wristband which tracks the child from time to time to ensure their safety. If any problem occurs it would alert parents through the cell phone so that they can take immediate action. This paper focus on the SMS text enabled communication. Parents can send SMS with some keywords and the device reply back. The device can detect the child's approximate location, it can detect the body temperature and the surrounding temperature, humidity and also the heartbeat of a child. For the emergency situation, the device would have some measures like an alarm buzzer, SOS light which will notify the bystanders to help the child. So this paper is all about the safety and security of a child to help them to recover from any type of difficulty.
{"title":"Multi-sensor Wearable for Child Safety","authors":"Ushashi Chowdhury, Pranjal Chowdhury, Sourav Paul, Anwesha Sen, Partho Protim Sarkar, S. Basak, Abari Bhattacharya","doi":"10.1109/UEMCON47517.2019.8992950","DOIUrl":"https://doi.org/10.1109/UEMCON47517.2019.8992950","url":null,"abstract":"Now-a-days we can see that human life is becoming very fast. Moreover, the city life is getting very busy day- by-day. So in the daily busy schedule it is becoming very difficult for the parents to monitor their children closely. This paper discusses about a smart wearable device like a wristband which tracks the child from time to time to ensure their safety. If any problem occurs it would alert parents through the cell phone so that they can take immediate action. This paper focus on the SMS text enabled communication. Parents can send SMS with some keywords and the device reply back. The device can detect the child's approximate location, it can detect the body temperature and the surrounding temperature, humidity and also the heartbeat of a child. For the emergency situation, the device would have some measures like an alarm buzzer, SOS light which will notify the bystanders to help the child. So this paper is all about the safety and security of a child to help them to recover from any type of difficulty.","PeriodicalId":187022,"journal":{"name":"2019 IEEE 10th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114220539","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：