Sound measurement and automatic vehicle classification and counting applied to road traffic noise characterization

Abstract

Increase in population density in large cities has increased the environmental noise present in these environments, causing negative effects on human health. There are different sources of environmental noise; however, noise from road traffic is the most prevalent in cities. Therefore, it is necessary to have tools that allow noise characterization to establish strategies that permit obtaining levels that do not affect the quality of life of people. This research discusses the implementation of a system that allows the acquisition of data to characterize the noise generated by road traffic. First, the methodology for obtaining acoustic indicators with an electret measurement microphone is described, so that it adjusts to the data collection needs for road traffic noise analyses. Then, an approach for the classification and counting of automatic vehicular traffic through deep learning is presented. Results showed that there were differences of 0.2 dBA in terms of RMSE between a type 1 sound level meter and the measurement microphone used. With reference to vehicle classification and counting for four categories, the approximate error is between 3.3% and -15.5%.