Recently, monitoring of road surface is a key factor for road maintenance and management. With the advances in the optical methods, the road monitoring systems have been equipped with high accuracy and resolution sensor package. However, most of the existing sensor packages are equipped with expensive equipment such as optical and complex sensors in considering the dynamics of mobile vehicles and dynamic outdoor environments. In this paper, we propose a CNN-based line laser refinement. The proposed system is designed based on the improvement of CNN-based line lasers, and it is more cost-effective than the existing expensive system.
{"title":"Road Surface Profiling based on Artificial-Neural Networks","authors":"Seungho Choi, Seoyeon Kim, Heelim Hong, Y. B. Kim","doi":"10.1145/3400286.3418282","DOIUrl":"https://doi.org/10.1145/3400286.3418282","url":null,"abstract":"Recently, monitoring of road surface is a key factor for road maintenance and management. With the advances in the optical methods, the road monitoring systems have been equipped with high accuracy and resolution sensor package. However, most of the existing sensor packages are equipped with expensive equipment such as optical and complex sensors in considering the dynamics of mobile vehicles and dynamic outdoor environments. In this paper, we propose a CNN-based line laser refinement. The proposed system is designed based on the improvement of CNN-based line lasers, and it is more cost-effective than the existing expensive system.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114901500","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}
Kicheol Park, Y. Lee, Jiman Hong, J. An, Bongjae Kim
With the rapid development of the Internet of Things (IoT) and AI technology, IoT services based on Artificial Intelligence (AI) technology are becoming more and more intelligent. To provide these intelligent IoT services, IoT hardware and IoT software must support AI technology. In general, battery-powered IoT devices have limited computing power compared to general-purpose computers. Therefore, to implement various intelligent IoT services, it must be able to support AI technology with low power to IoT devices. The low-power Neuromorphic architecture can enable resource-limited IoT devices to provide intelligent IoT services based on AI technology. In this paper, we propose a Neuromorphic Architecture Abstraction (NAA) model for providing an efficient intelligent IoT service. The proposed NAA model dynamically selects the proper Neuromorphic architecture according to the characteristics of the training target architecture and increases the training speed and training success rate. We also implement the proposed model in a real IoT computing environment and show that the proposed NAA model can reduce the training speed and reduce the training models success rate compared with the method of randomly specifying the Neuromorphic architecture.
{"title":"Selecting a Proper Neuromorphic Platform for the Intelligent IoT","authors":"Kicheol Park, Y. Lee, Jiman Hong, J. An, Bongjae Kim","doi":"10.1145/3400286.3418264","DOIUrl":"https://doi.org/10.1145/3400286.3418264","url":null,"abstract":"With the rapid development of the Internet of Things (IoT) and AI technology, IoT services based on Artificial Intelligence (AI) technology are becoming more and more intelligent. To provide these intelligent IoT services, IoT hardware and IoT software must support AI technology. In general, battery-powered IoT devices have limited computing power compared to general-purpose computers. Therefore, to implement various intelligent IoT services, it must be able to support AI technology with low power to IoT devices. The low-power Neuromorphic architecture can enable resource-limited IoT devices to provide intelligent IoT services based on AI technology. In this paper, we propose a Neuromorphic Architecture Abstraction (NAA) model for providing an efficient intelligent IoT service. The proposed NAA model dynamically selects the proper Neuromorphic architecture according to the characteristics of the training target architecture and increases the training speed and training success rate. We also implement the proposed model in a real IoT computing environment and show that the proposed NAA model can reduce the training speed and reduce the training models success rate compared with the method of randomly specifying the Neuromorphic architecture.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130487030","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}
Yung-Feng Lu, Hung-Ming Chen, Chin-Fu Kuo, Bo-Ting Chen, Zong-Yan Dai
The core concept of electronic money is blockchain, which can be regarded as a decentralized database. It is a decentralized storage service that does not rely on third-party storage. It records each transaction information on the block. Related applications include electronic ledgers, which further extend smart contracts. In storage, not only record transactions, but the extended smart contract can be used in medical treatment, recording the patient's medical records. However, there is a need for privacy in medical records, and it must be able to be accessed by appropriate people in accordance with the authority. So it needs a technology like Enigma to achieve it. Enigma uses multi-party computation (hereinafter referred to as MPC) and distributed hash-table (hereinafter referred to as DHT) technology, combined with the blockchain to divide the data to be stored into public blocks and private parts The blockchain has an existing way of computing and storing, and private data is Enigma's method of computing and storing by using the method called Off-Chain. Only authenticated users can access private data. However, many applications have the need for group sharing and decentralized records. This study uses Enigma to encrypt the data that requires privacy using the key of the encrypted file and then uploads it to DHT for storage, so that the owner of the private data can further share the data securely.
{"title":"Enhanced Privacy with Blockchain-based Storage for Data Sharing","authors":"Yung-Feng Lu, Hung-Ming Chen, Chin-Fu Kuo, Bo-Ting Chen, Zong-Yan Dai","doi":"10.1145/3400286.3418242","DOIUrl":"https://doi.org/10.1145/3400286.3418242","url":null,"abstract":"The core concept of electronic money is blockchain, which can be regarded as a decentralized database. It is a decentralized storage service that does not rely on third-party storage. It records each transaction information on the block. Related applications include electronic ledgers, which further extend smart contracts. In storage, not only record transactions, but the extended smart contract can be used in medical treatment, recording the patient's medical records. However, there is a need for privacy in medical records, and it must be able to be accessed by appropriate people in accordance with the authority. So it needs a technology like Enigma to achieve it. Enigma uses multi-party computation (hereinafter referred to as MPC) and distributed hash-table (hereinafter referred to as DHT) technology, combined with the blockchain to divide the data to be stored into public blocks and private parts The blockchain has an existing way of computing and storing, and private data is Enigma's method of computing and storing by using the method called Off-Chain. Only authenticated users can access private data. However, many applications have the need for group sharing and decentralized records. This study uses Enigma to encrypt the data that requires privacy using the key of the encrypted file and then uploads it to DHT for storage, so that the owner of the private data can further share the data securely.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132947500","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}
Asymmetric multicore architecture is widely applied to the embedded systems to better trade-off performance and energy consumption. With an increased number of applications concurrently executed in the system, the power consumption and the associated last-level cache latency are increased. To maximize the system performance under the power constraint, we proposed a cache contention-aware run-time scheduling for asymmetric multicore systems. To deal with the dynamic workload and cache contention effect, the CPI model learning is presented to adjust the relation between system performance, executing frequency, and executing clusters. Based on the CPI model prediction, the run-time dispatcher is then presented to determine the executing frequency and cores to maximize system throughput under power constraint. The proposed algorithm was implemented on the commercial Odroid XU4 board. The performance was evaluated using benchmarks and impressive results were obtained.
{"title":"A Cache Contention-aware Run-time Scheduling for Power-constrained Asymmetric Multicore Processors","authors":"Jian-He Liao, He-Ru Chen, Ya-Shu Chen","doi":"10.1145/3400286.3418230","DOIUrl":"https://doi.org/10.1145/3400286.3418230","url":null,"abstract":"Asymmetric multicore architecture is widely applied to the embedded systems to better trade-off performance and energy consumption. With an increased number of applications concurrently executed in the system, the power consumption and the associated last-level cache latency are increased. To maximize the system performance under the power constraint, we proposed a cache contention-aware run-time scheduling for asymmetric multicore systems. To deal with the dynamic workload and cache contention effect, the CPI model learning is presented to adjust the relation between system performance, executing frequency, and executing clusters. Based on the CPI model prediction, the run-time dispatcher is then presented to determine the executing frequency and cores to maximize system throughput under power constraint. The proposed algorithm was implemented on the commercial Odroid XU4 board. The performance was evaluated using benchmarks and impressive results were obtained.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127092076","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}
Dipta Das, Micah Schiewe, Elizabeth Brighton, Mark Fuller, T. Cerný, Miroslav Bures, Karel Frajták, Dongwan Shin, Pavel Tisnovsky
In modern computing, log files provide a wealth of information regarding the past of a system, including the system failures and security breaches that cost companies and developers a fortune in both time and money. While this information can be used to attempt to recover from a problem, such an approach merely mitigates the damage that has already been done. Detecting problems, however, is not the only information that can be gathered from log files. It is common knowledge that segments of log files, if analyzed correctly, can yield a good idea of what the system is likely going to do next in real-time, allowing a system to take corrective action before any negative actions occur. In this paper, the authors put forth a systematic map of this field of log prediction, screening several hundred papers and finally narrowing down the field to approximately 30 relevant papers. These papers, when broken down, give a good idea of the state of the art, methodologies employed, and future challenges that still must be overcome. Findings and conclusions of this study can be applied to a variety of software systems and components, including classical software systems, as well as software parts of control, or the Internet of Things (IoT) systems.
{"title":"Failure Prediction by Utilizing Log Analysis: A Systematic Mapping Study","authors":"Dipta Das, Micah Schiewe, Elizabeth Brighton, Mark Fuller, T. Cerný, Miroslav Bures, Karel Frajták, Dongwan Shin, Pavel Tisnovsky","doi":"10.1145/3400286.3418263","DOIUrl":"https://doi.org/10.1145/3400286.3418263","url":null,"abstract":"In modern computing, log files provide a wealth of information regarding the past of a system, including the system failures and security breaches that cost companies and developers a fortune in both time and money. While this information can be used to attempt to recover from a problem, such an approach merely mitigates the damage that has already been done. Detecting problems, however, is not the only information that can be gathered from log files. It is common knowledge that segments of log files, if analyzed correctly, can yield a good idea of what the system is likely going to do next in real-time, allowing a system to take corrective action before any negative actions occur. In this paper, the authors put forth a systematic map of this field of log prediction, screening several hundred papers and finally narrowing down the field to approximately 30 relevant papers. These papers, when broken down, give a good idea of the state of the art, methodologies employed, and future challenges that still must be overcome. Findings and conclusions of this study can be applied to a variety of software systems and components, including classical software systems, as well as software parts of control, or the Internet of Things (IoT) systems.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114782388","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}
Kwanghee Won, Youjeong Jang, Hyung-Do Choi, Sung Y. Shin
Semantic classification of scientific literature using machine learning approaches is challenging due to the lack of labeled data and the length of text [1, 4]. Most of the work has been done for keyword based categorization tasks, which take care of occurrence of important terms, whereas the semantic classification is to learn keywords as well as the meaning of sentences. In this study, we have evaluated neural network models on a semantic classification task using a large amount of labeled scientific papers listed in the Powerwatch study. We have conducted neural architecture search to find the most suitable model for the task. In the experiment, we have compared classification accuracy of various neural network models. In addition, we have employed a Fully Convolutional Neural Network (FCN) to implement attention mechanism for the semantic classification of EMF-related literature. The experimental result showed that the FCN-based attention model was able to identify important parts of input texts.
{"title":"Semantic Classification of EMF-related Literature using Deep Learning Models with Attention Mechanism","authors":"Kwanghee Won, Youjeong Jang, Hyung-Do Choi, Sung Y. Shin","doi":"10.1145/3400286.3418259","DOIUrl":"https://doi.org/10.1145/3400286.3418259","url":null,"abstract":"Semantic classification of scientific literature using machine learning approaches is challenging due to the lack of labeled data and the length of text [1, 4]. Most of the work has been done for keyword based categorization tasks, which take care of occurrence of important terms, whereas the semantic classification is to learn keywords as well as the meaning of sentences. In this study, we have evaluated neural network models on a semantic classification task using a large amount of labeled scientific papers listed in the Powerwatch study. We have conducted neural architecture search to find the most suitable model for the task. In the experiment, we have compared classification accuracy of various neural network models. In addition, we have employed a Fully Convolutional Neural Network (FCN) to implement attention mechanism for the semantic classification of EMF-related literature. The experimental result showed that the FCN-based attention model was able to identify important parts of input texts.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122386132","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}
The Bloom filter is a hash-based data structure that facilitates membership querying. Computation speed of Bloom filter is affected by hash functions that produce hash outputs. Basically, two operations: 'add' and 'query', consists of the Bloom filter. Previous researches have shown advanced computation speed of Bloom filter since the standard Bloom Filter is published. For example, Double Hash Bloom filter, Single Hash Bloom filter, etc. We propose a system that uses less computation than previous works with similar false positive and which is implemented by using a circular bit shift. This method was implemented with faster calculation speed, compared with previous works. Furthermore, experiments which were compared with previous researches and standard Bloom filter. Therefore, we demonstrate that the proposed system computes faster than previous studies with similar false positive rate.
{"title":"Modifications using Circular Shift for a Better Bloom Filter","authors":"Myeonghun Kim, Sung-Ryul Kim","doi":"10.1145/3400286.3418232","DOIUrl":"https://doi.org/10.1145/3400286.3418232","url":null,"abstract":"The Bloom filter is a hash-based data structure that facilitates membership querying. Computation speed of Bloom filter is affected by hash functions that produce hash outputs. Basically, two operations: 'add' and 'query', consists of the Bloom filter. Previous researches have shown advanced computation speed of Bloom filter since the standard Bloom Filter is published. For example, Double Hash Bloom filter, Single Hash Bloom filter, etc. We propose a system that uses less computation than previous works with similar false positive and which is implemented by using a circular bit shift. This method was implemented with faster calculation speed, compared with previous works. Furthermore, experiments which were compared with previous researches and standard Bloom filter. Therefore, we demonstrate that the proposed system computes faster than previous studies with similar false positive rate.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123434099","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}
Min-yu Tsai, Y. Lai, Y. Chi, X. Jia, Shih-Hao Hung
The simulation of water radiolysis including three stages, physical, physico-chemical and chemical, modeling the interactions between water and radicals is essential to understand the radiobiological mechanisms and quantitatively test some hypotheses in related problem. Monte Carlo (MC) simulation is recognized as one of the most accurate approaches for the computations of the water radiolysis process. Geant4-DNA which extending the Geant4 Monte Carlo simulation toolkit provides accurate descriptions of the initial physical process of ionization, along with the pre-chemical production of ion species and subsequent chemistry, in a single application for water radiolysis. To accelerate the long execution time of Geant4-DNA simulation, an open source GPU code for water radiolysis simulation, gMicroMC, has been developed. In this paper, we focus on reviewing the GPU implementation architecture of each stage of gMicroMC and evaluating the computational performance in the sub-MeV range of incident electrons. The experimental results of gMicroMC show up to three orders of magnitude performance gain, up to 1690x, with recent generations of NVIDIA graphic cards compared with Geant4-DNA running on a single CPU thread.
{"title":"Performance Evaluation of a GPU-based Monte Carlo Simulation Package for Water Radiolysis with sub-MeV Electrons","authors":"Min-yu Tsai, Y. Lai, Y. Chi, X. Jia, Shih-Hao Hung","doi":"10.1145/3400286.3418241","DOIUrl":"https://doi.org/10.1145/3400286.3418241","url":null,"abstract":"The simulation of water radiolysis including three stages, physical, physico-chemical and chemical, modeling the interactions between water and radicals is essential to understand the radiobiological mechanisms and quantitatively test some hypotheses in related problem. Monte Carlo (MC) simulation is recognized as one of the most accurate approaches for the computations of the water radiolysis process. Geant4-DNA which extending the Geant4 Monte Carlo simulation toolkit provides accurate descriptions of the initial physical process of ionization, along with the pre-chemical production of ion species and subsequent chemistry, in a single application for water radiolysis. To accelerate the long execution time of Geant4-DNA simulation, an open source GPU code for water radiolysis simulation, gMicroMC, has been developed. In this paper, we focus on reviewing the GPU implementation architecture of each stage of gMicroMC and evaluating the computational performance in the sub-MeV range of incident electrons. The experimental results of gMicroMC show up to three orders of magnitude performance gain, up to 1690x, with recent generations of NVIDIA graphic cards compared with Geant4-DNA running on a single CPU thread.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120849618","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}
Mobile applications are nowadays one of the most important parts in every person's life. Today's smartphones are more capable of operating different type of applications. Mobile devices can create an interface between people and service providers. But due to limited computation power, storage, and battery capacity, it is not possible to computing all process in mobile devices. Thus, a cloud-based solution using mobile devices is an effective solution to overcome smartphones' constraints and gets access through anywhere for different kinds of facilities and services by offloading computation activities from mobile devices. These outputs of that computation are returned back to mobile devices. However, cloud resources are also limited. In order to use the limited cloud resources, load-balancing solutions are recommended for improving Hybrid cloud computing architectures. In this paper, we address why hybrid cloud computing is important in mobile applications and also demonstrate why load balancing is required in hybrid cloud computing architectures. We demonstrate and evaluate load-balancing performances using the ns-3 network simulation tool.
{"title":"A Study of Load-Balancing Solutions of Mobile Cloud Computing for Next-Generation Mobile Applications","authors":"Rupak Kumar Das, Ahyoung Lee","doi":"10.1145/3400286.3418238","DOIUrl":"https://doi.org/10.1145/3400286.3418238","url":null,"abstract":"Mobile applications are nowadays one of the most important parts in every person's life. Today's smartphones are more capable of operating different type of applications. Mobile devices can create an interface between people and service providers. But due to limited computation power, storage, and battery capacity, it is not possible to computing all process in mobile devices. Thus, a cloud-based solution using mobile devices is an effective solution to overcome smartphones' constraints and gets access through anywhere for different kinds of facilities and services by offloading computation activities from mobile devices. These outputs of that computation are returned back to mobile devices. However, cloud resources are also limited. In order to use the limited cloud resources, load-balancing solutions are recommended for improving Hybrid cloud computing architectures. In this paper, we address why hybrid cloud computing is important in mobile applications and also demonstrate why load balancing is required in hybrid cloud computing architectures. We demonstrate and evaluate load-balancing performances using the ns-3 network simulation tool.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134243698","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}
The usage of quadcopter types of drones is now on mature and a practical stage and many major manufacturers are expanding its applications into various regions with it. Considerable characteristic of this type of flying object as its maneuverability and practicality is now being focused on how we control this among our urban life from the possibility of any offensive usage. Most of them are either small enough to avoid many current airborne detection methods and cheap enough to use them as disposable. In this paper, we tried to analyze the recorded sounds of a subset of commercial quadcopter types of drones and built a trained simple non-linear neural network filter to classify them among the given sound samples. We borrowed Mel-frequency cepstral coefficients as the well-known methodology of sound analysis process but including some of the parameter adjustments for this research, and applied LeNet neural network filter structure for the following classification test. To maintain the information of adjacent samples among the series of wave samples, 2-D spectrogram planning was applied as for the input signal preprocessing. Most of the frequencies from drones were observed as gathered around 3 to 5Khz, up to around 10Khz, and adjusted LeNet architecture could classify over 10 types of drone categories with over 95% of accuracy.
{"title":"Analysis of commercial drone sounds and its identification","authors":"Sinwoo Yoo, H. Oh","doi":"10.1145/3400286.3418267","DOIUrl":"https://doi.org/10.1145/3400286.3418267","url":null,"abstract":"The usage of quadcopter types of drones is now on mature and a practical stage and many major manufacturers are expanding its applications into various regions with it. Considerable characteristic of this type of flying object as its maneuverability and practicality is now being focused on how we control this among our urban life from the possibility of any offensive usage. Most of them are either small enough to avoid many current airborne detection methods and cheap enough to use them as disposable. In this paper, we tried to analyze the recorded sounds of a subset of commercial quadcopter types of drones and built a trained simple non-linear neural network filter to classify them among the given sound samples. We borrowed Mel-frequency cepstral coefficients as the well-known methodology of sound analysis process but including some of the parameter adjustments for this research, and applied LeNet neural network filter structure for the following classification test. To maintain the information of adjacent samples among the series of wave samples, 2-D spectrogram planning was applied as for the input signal preprocessing. Most of the frequencies from drones were observed as gathered around 3 to 5Khz, up to around 10Khz, and adjusted LeNet architecture could classify over 10 types of drone categories with over 95% of accuracy.","PeriodicalId":326100,"journal":{"name":"Proceedings of the International Conference on Research in Adaptive and Convergent Systems","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130414073","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：