Noise levels of anthropogenic origin in urban environments have reached thresholds that pose serious public health and quality of life problems. This paper/work aims to examine these noise levels, the underlying causes of their increase and possible solutions through the implementation of predictive models. To address this problem, as a first step, a simplified mathematical model capable of accurately predicting anthropogenic noise levels in a given area is developed. As variables, this model considers the crowd density, estimated using an Artificial Neural Network (ANN) capable of detecting people in images, as well as the geometric and architectural characteristics of the environment. To verify the model, several protocols have been developed for collecting experimental data. In a first phase, these experimental measurements were carried out in controlled environments, using loudspeakers as noise sources. In a second phase, these measurements were carried out in real environments, accounting for the specific noise sources present in each setting. The difference in sound levels between the model and reality is proven to be less than 3 dB in 75% and less than 3.5 dB in 100% of the cases examined in a controlled environment. In the real problem, in general terms and taking into account that the study is carried out on pedestrian streets, it seems that the model is able to reproduce most of the noise of anthropogenic origin.