Loudspeakers as recording devices in public address systems

Abstract

Public address (PA) systems are an integral part of every building. Their function as a safety and security system could benefit from adding recording capabilities. With such capabilities, people could be located during evacuation or announcements could be adjusted to room acoustics for better intelligibility. However, adding microphones to a PA system is too costly and inefficient to be feasible. This thesis evaluates the feasibility of using the loudspeakers already in place as recording devices to provide information to the system. To this end, a system using a single loudspeaker as both a playback and recording device is analysed, modelled and simulated. The results show that using a current measuring set-up with an analogue-to-digital converter capable of detecting a range of roughly 120 dB, a speech signal up to three metres in a cone of 120° from a loudspeaker can be successfully estimated in an office room with an announcement playing and background noise present. As the estimated signal is unknown to the system, the solution generalises to other signal types as well. A system with a single loudspeaker can be utilised for the use cases presented, and is therefore proven feasible. To increase the practicality of the system, it is recommended to continue the research in two main areas. The first area considers improving the quality of the recording and extending the range from which the system can produce accurate recordings. The second area considers evaluating the practical implementation of the system, by extending the single loudspeaker case to a multiple loudspeaker case and generalising the system to be readily implementable in a large variety of PA systems.