Development of an Artificial Intelligence for Efficient Noise Management in Construction Sites: Enhancing the Quality of Environmental Data and the Sound Source Classification Model

• Grant number: 24/18391-0
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: May 01, 2025
• End date: January 31, 2026
• Field of knowledge: Interdisciplinary Subjects
• Principal Investigator: Carolina Rodrigues Alves Monteiro
• Grantee: Carolina Rodrigues Alves Monteiro
• Company: OTOH LTDA
• CNAE: Desenvolvimento de programas de computador sob encomenda ; Desenvolvimento e licenciamento de programas de computador não-customizáveis ; Testes e análises técnicas
• City: São Paulo
• Associated researchers: Cecilia Jardim Gomes ; Leonardo Jacomussi Pereira de Araujo ; Marcel Pozzobon Borin
• Associated scholarship(s): 25/07077-6 - Development of an artificial intelligence for efficient noise management in construction sites: enhancing the quality of environmental data and the sound source classification model, BP.PIPE

Abstract

Currently, OTOH - a company specializing in noise monitoring with remote sensing and data access - uses embedded artificial intelligence (edge AI) to identify urban sound sources. Developed with neural networks and an open-source audio database, the AI recognizes some generic aspects of the soundscape, such as air and road traffic, animals, and human sounds. However, the lack of contextualization and the low sampling of construction noise causes errors in acoustic analysis. Regarding data access, it is necessary to improve the way users interact beyond conventional data visualization (with graphs and tables), accelerating the interpretation of results and enabling more assertive actions in construction sites related to noise management.To address these challenges, we propose to optimize the collection and classification of audio samples on-site and develop a generative AI agent that will provide noise management suggestions (insights). Its training will include static information relevant to the context (fine-tuning), such as legislation addressing environmental noise issues that set sound level limits according to time of day and/or urban zoning, construction schedules, public health guidelines/indicators, and will have access to dynamic information (RAG) provided in real-time by the OTOH IoT system.To achieve the objective, the following actions will be necessary: expanding the audio database focusing on sound sources from construction sites to improve the accuracy of the source identification model; synthesizing a set of parameters sufficient to contextualize the soundscape of a construction site, thereby automating the collection of relevant information to train an LLM model and create specialized agents for each project.By the end of the research project, the developed AI algorithm is expected to generate relevant suggestions for noise management in construction sites, thus reducing the impact of noise pollution in the urban soundscape. (AU)