Direction of arrival estimation of vehicle sound sources in a Two-Microphone Array

• Ano de defesa: 2020
• Autor(a) principal: Rocha, Gabriela Dantas
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Elétrica; UFRJ
• Palavras-chave em Português: Direção de chegada; Ruído do trânsito; Estimativa de distância entre eixos; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA
• Link de acesso: http://hdl.handle.net/11422/25642

Conventional methods for the direction of arrival estimation of sound sources are capable of generating accurate estimates in applications with only one source. In the context of traffic noise, each vehicle has at least two important sound sources, associated with the tires on the front and rear axles. An extension of methods for the direction of arrival estimation of sound sources from a set of two microphones is proposed in this work, in order to deal with multiple sources. The proposed method is divided into two stages: in the first, one of the traditional arrival direction estimation algorithms is used to generate the delay matrix; in the second, the delay matrix is treated, using image processing techniques, and then used as an input to a curve fitting algorithm. Four methods are tested for generating the delay matrix: Generalized Cross-Correlation (GCC), Interaural Time Difference (ITD), Least Mean Square (LMS) and Eigenvalue Decomposition (EVD). Two curves are obtained as a result of the second stage, representing the time difference of arrival estimated of the sounds emitted by the car tires on the two microphones over time. In addition to the directions of arrival of the two main sound sources, estimates of the speed and distance between the axles of the cars can be obtained from the two curves. Measurements of the sound emitted during the individual passage of cars of four models at different speeds are used to test the performance of the system. The experimental results indicate that, with the modifications introduced in the second stage, the system was able to estimate the distance between the axles of the cars with an average error of 26 cm using the GCC algorithm, and to estimate the speed with an average error of 10.01 km/h using the LMS algorithm.