This project investigates the design of a microphone array configuration to localise acoustic sources in air.
One situation in particular is the case of multiple concurrent sound sources both within dry and reverberant environments. Previously many sound localisation techniques have been developed around microphone arrays which estimate the acoustic intensity vector of a source by measuring the pressure difference between spatially separated omnidirectional microphones. This technique does not work well when multiple concurrent sound sources are present. The techniques explored in this research use novel acoustic vector sensor type microphones which measure the acoustic intensity vector via one pressure sensor and three differential microphones, all collocated. The performance of these microphones will be evaluated and will then be used as the base upon which an acoustic vector array configuration will be designed. The performance of this array will be measured by how well it can localise multiple sources within different environments. The same microphone array can also be used to beamform to a particular sound source, i.e., to acoustically separate the source from surrounding noise. A key application of this research is multi-speaker voice transcription (such as in a teleconference situation) where the key speaker in a room needs to be found (localisation) and then separated from opposing speakers/noise (beamforming) so that an accurate transcription can be generated by automatic speech recognition software.
The opportunity ID for this research opportunity is 273