Anti-Mosquito System I: identification and localization of acoustic sources

• Autor(a) principal: Reis, José Alberto Barros dos
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/35126
• topic: Mosquitoes; Trajectory; Microphones; Angle of arrival; Time difference of arrival; Cross-correlation; Azimuth; Far-field; Real-time; Room simulation

This dissertation addresses the development of an acoustic localisation system with the aim of detecting mosquitoes indoors. It starts with a brief study of the sound produced by insects, with special focus on the case of female mosquitoes, aimed at understanding the spectral characteristics; A review was carried out on our auditory system and its ability to spatially locate sound sources. The main 2D cues are ITD (interaural time difference) and ILD (interaural level difference). The example of human hearing shows how spatial diversity of sensors is indispensable for sound localisation; A 2D scenario was assumed, thus reducing the problem to azimuth estimation, which requires two microphones. Assuming that the distance from the source to the receiver is much greater than the distance between microphones (far-field approximation) the sought azimuth angle can be obtained by an approximate formula. The intrinsic error caused by the far-field approximation itself was assessed, as well as the impact of possible estimation errors in the calculation parameters: speed of sound, microphone spacing and time delay; The development work, carried out on a MATLAB environment, was based on an existing simulator. The central element of the system is the digital processing of the signals received at the two microphones. The cross-correlation method is used to work out the time delay between them. Interpolation was applied to increase the resolution of the cross-correlation peak estimate; A script featuring a graphical interface was developed to combine the predictor with the simulator. It makes it easy for the user to specify the trajectory to be reproduced in the simulator. The audio file to be injected is also chosen by the user. The simulator returns a stereo file with the microphone signals. The script generates a pointer moving in real time to indicate the estimated position of the source; Several other simulations and experimental tests were carried out, based on an anechoic room without additional sources of noise. The azimuth estimation error measured in simulation confirmed the predicted behaviour taking into account the sources of error intrinsic to the far-field approximation. The error is smaller when the source is between 45° and 135°. Outside this range, it increases, peaking at the extremes (0° and 180°). It approaches zero when the source is at 90°, forming a symmetric U-shaped pattern around this value. When noise is introduced, the estimations made lose quality, as expected; for SNR less than -10 dB, the error exceeds 10°; The experimental tests involved two microphones, a loudspeaker and an audio interface for communication with the computer. An absorbing chamber has been created to reduce sound reflections and external noise. Recordings of long duration were made for each azimuth angle. With all the files processed, the pattern of the azimuth estimation error was also U-shaped, although not perfectly symmetric.