Machine learning techniques applied to radio frequency (RF) signals are used for many applications in addition to data communication. In this paper, the authors propose a machine learning solution for classifying the number of people within an indoor ambient. The main idea is to identify a pattern of received signal characteristics according to the number of people. Experimental measurements are performed using a software-defined radio platform inside a laboratory. The data collected is post-processed by applying a feature mapping technique based on mean, standard deviation, and Shannon information entropy. This feature-space data is then used to train a supervised machine learning network for classifying scenarios with zero, one, two, and three people inside. The proposed solution presents significant accuracy in classification performance.

Resumo-Com a definição pelo ITU dos requisitos para os sistemas 5G por meio de uma série de documentos denominados IMT for 2020 and beyond, os sistemas 5G mMTC (massive Machine Type Communications) foram incluídos como um dos três use cases para próxima geração de sistemas de comunicações móveis. Para tais sistemas, a cobertura extensa (15 km) e um dos tripés de desempenho, podendo ser viabilizada ao operar em faixas de frequência em torno de centenas de MHz (700-900 MHz). Assim, caracterizar/modelar o canal em tais faixas se torna crucial para fomentar a concepção de funcionalidades para os sistemas 5G mMTC. Este artigo tem como objetivo a apresentação de um sistema de baixo custo para medições e caracterização de canal em 700 MHz. A Universal Software Radio Peripheral (USRP) foi usada como plataforma de hardware em conjunto com o software GNU Radio. O pós-processamento das medições foi realizado na plataforma de software livre R, a qual disponibiliza uma série de ferramentas estatísticas e gráficas. A caracterização em banda estreita do canalé apresentada no cenário Dual-stripe, definido pelo 3GPP.Palavras-Chave-5G mMTC, Dual-stripe indoor, Canal, GNU Radio, USRP.Abstract-With ITU's definition of requirements for 5G systems through a series of documents called IMT for 2020 and beyond, 5G mMTC (massive Machine Type Communications) systems were included as one of three use cases for next generation of mobile communications systems. For such systems, extensive coverage (15 km) is one of the performance tripods, and it can be made possible by operating in frequency bands around hundreds of MHz (700-900MHz). Thus, characterizing/modeling the channel in such bands becomes crucial to help the design of functionalities for 5G mMTC systems. This paper has as main objective the presentation of a low cost system for measurements and channel characterization at 700 MHz. Universal Software Radio Peripheral (USRP) is used as the hardware platform in conjunction with the GNU Radio software. The measurements post-processing is performed in the free software platform R, which provides a series of statistical and graphical tools. The narrowband characterization of the channel is presented in the Dual-stripe scenario, as defined by 3GPP.