Implementation of an Autonomous Impulse Response Measurement System

Abstract

Data collection is crucial for researchers, as it can provide important insights for describing phenomena. In acoustics, acoustic phenomena are characterized by Room Impulse Responses (RIRs) occurring when sound propagates in a room. Room impulse responses are needed in vast quantities for various reasons, including the prediction of acoustical parameters and the rendering of virtual acoustical spaces. Recently, mobile robots navigating within indoor spaces have become increasingly used to acquire information about its environment. However, little research has attempted to utilize robots for the collection of room acoustic data.

This thesis presents an adaptable automated system to measure room impulse responses in multi-room environments, using mobile and stationary measurement platforms. The system, known as Autonomous Impulse Response Measurement System (AIRMS), is divided into two stages: data collection and post-processing. To automate data collection, a mobile robotic platform was developed to perform acoustic measurements within a room. The robot was equipped with spatial microphones, multiple loudspeakers and an indoor localization system, which reported real time location of the robot. Additionally, stationary platforms were installed in specific locations inside and outside the room. The mobile and stationary platforms wirelessly communicated with one another to perform the acoustical tests systematically. Since a major requirement of the system is adaptability, researchers can define the elements of the system according to their needs, including the mounted equipment and the number of platforms. Post-processing included extraction of sine sweeps and the calculation of impulse responses. Extraction of the sine sweeps refers to the process of framing every acoustical test signal from the raw recordings. These signals are then processed to calculate the room impulse responses. The automatically collected information was complemented with manually produced data, which included rendering of a 3D model of the room, a panoramic picture.

The performance of the system was tested under two conditions: a single-room and a multiroom setting. Room impulse responses were calculated for each of the test conditions, representing typical characteristics of the signals and showing the effects of proximity from sources and receivers, as well as the presence of boundaries. This prototype produces RIR measurements in a fast and reliable manner.

Although some shortcomings were noted in the compact loudspeakers used to produce the sine sweeps and the accuracy of the indoor localization system, the proposed autonomous measurement system yielded reasonable results. Future work could expand the amount of impulse response measurements in order to further refine the artificial intelligence algorithms.