Evaluation of room-acoustic modal characteristics from single-point measurements using Bayesian analysis

Abstract

Room mode analysis is an important element of architectural acoustic design. In small rooms especially, well-separated low frequency room modes can cause unpleasant aural effects, such as undesirable resonances and flutter echoes. Traditional room mode analysis is generally done using the discrete Fourier transform. This work proposes a time-domain modal analysis method by which the amplitudes, frequencies, and damping constants of the room under test can be directly determined for low modal frequencies by using a Bayesian inference algorithm on a single room impulse response (RIR). The method’s time-domain, model-based approach allows the number of modes present in the RIR, as well as the amplitudes, frequencies, and damping constants for each mode, to be determined using Bayesian model selection and parameter estimation, respectively. The method uses Skilling’s nested sampling algorithm to infer parameter values. Results indicate that the method is especially useful in rooms with closely-spaced modes at low frequencies.