Constructing a noise pollution map for an area includes a first subset of users performing initial noise recordings in the area using audio devices, using machine learning to provide classification of noises in the initial noise recordings, a second subset of users, larger than the first subset of users, capturing noise in the area using audio devices, creating summaries of noises using the classification to classify noises captured by the second subset of users, and aggregating the summaries to construct the noise pollution map of the area. The audio devices may include headsets, smart speakers, smart television sets, and/or computers. The summaries of noises may be created using software that is installed locally on devices of the second subset of users. The summaries may include source information, amplitude and frequency characteristics, duration, parameters of a corresponding one of the audio devices, user location, surroundings, and/or user movement information.