George D. O’Mahony, Philip J. Harris, Colin C. Murphy
{"title":"Investigating Supervised Machine Learning Techniques for Channel Identification in Wireless Sensor Networks","authors":"George D. O’Mahony, Philip J. Harris, Colin C. Murphy","doi":"10.1109/ISSC49989.2020.9180209","DOIUrl":null,"url":null,"abstract":"Knowledge of the wireless channel is pivotal for wireless communication links but varies for multiple reasons. The radio spectrum changes due to the number of connected devices, demand, packet size or services in operation, while fading levels, obstacles, path losses, and spurious (non-)malicious interference fluctuate in the physical environment. Typically, these channels are applicable to the time series class of data science problems, as the primary data points are measured over a period. In the case of wireless sensor networks, which regularly provide the device to access point communication links in Internet of Things applications, determining the wireless channel in operation permits channel access. Generally, a clear channel assessment is performed to determine whether a wireless transmission can be executed, which is an approach containing limitations. In this study, received in-phase (I) and quadrature-phase (Q) samples are collected from the wireless channel using a software-defined radio (SDR) based procedure and directly analyzed using python and Matlab. Features are extracted from the probability density function and statistical analysis of the received I/Q samples and used as the training data for the two chosen machine learning methods. Data is collected and produced over wires, to avoid interfering with other networks, using SDRs and Raspberry Pi embedded devices, which utilize available open-source libraries. Data is examined for the signal-free (noise), legitimate signal (ZigBee) and jamming signal (continuous wave) cases in a live laboratory environment. Support vector machine and Random Forest models are each designed and compared as channel identifiers for these signal types.","PeriodicalId":351013,"journal":{"name":"2020 31st Irish Signals and Systems Conference (ISSC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"3","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 31st Irish Signals and Systems Conference (ISSC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISSC49989.2020.9180209","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 3
Abstract
Knowledge of the wireless channel is pivotal for wireless communication links but varies for multiple reasons. The radio spectrum changes due to the number of connected devices, demand, packet size or services in operation, while fading levels, obstacles, path losses, and spurious (non-)malicious interference fluctuate in the physical environment. Typically, these channels are applicable to the time series class of data science problems, as the primary data points are measured over a period. In the case of wireless sensor networks, which regularly provide the device to access point communication links in Internet of Things applications, determining the wireless channel in operation permits channel access. Generally, a clear channel assessment is performed to determine whether a wireless transmission can be executed, which is an approach containing limitations. In this study, received in-phase (I) and quadrature-phase (Q) samples are collected from the wireless channel using a software-defined radio (SDR) based procedure and directly analyzed using python and Matlab. Features are extracted from the probability density function and statistical analysis of the received I/Q samples and used as the training data for the two chosen machine learning methods. Data is collected and produced over wires, to avoid interfering with other networks, using SDRs and Raspberry Pi embedded devices, which utilize available open-source libraries. Data is examined for the signal-free (noise), legitimate signal (ZigBee) and jamming signal (continuous wave) cases in a live laboratory environment. Support vector machine and Random Forest models are each designed and compared as channel identifiers for these signal types.