Abstract
When occupied by an audience, a musical performance space exhibits different acoustic characteristics compared to when it is empty. To obtain acoustical measurements with an audience present would require them to sit through an entire session of acoustic test signals, which is not practical. The dynamics of room acoustics parameters during an on-going concert may also be important to the mixing and recording engineers. These acoustic measurement scenarios demand an analysis tool for performing the designated tests during music concert sessions while not being noticeable by the audience. An acoustic measurement technique employing test signals generated during a musical performance is proposed. This method employs an adaptive time-frequency synthesis algorithm that determines the energy vacancies in the spectrogram of the on-going performance and automatically generates low-energy test signals that fill in the vacancies. The energy level of the test signal can be low enough to be masked by the music while high enough to be measurable above the noise floor. Several classes of test signals are proposed and their implementations in live musical events are demonstrated. Different system identification methods for estimating architectural acoustics parameters from the recorded test signals are also discussed and compared.