Bayesian model selection applied to room‐acoustic energy decay analysis

Abstract

The previous work of Xiang et al. [Xiang & Goggans, J. Acoust. Soc. Am. 110 (2001), pp. 1415‐1424; 113 (2003), pp. 2685‐2697; Xiang & Jasa, J. Acoust. Soc. Am. 120 (2006) pp.3744‐3749] successfully applied the Bayesian formulism to estimate multiple decay parameters from Schroeder decay functions measured or calculated in acoustically coupled spaces. In this work we consider a more difficult problem of determining the correct decay model in the presence of energy decay data within the Bayesian framework. We will compare the Annealing/Thermodynamic algorithm [Neal, R. M., Statistics and Computing 11 (2001), pp. 125‐139], the Nested Sampling algorithm [Silvia & Skilling, Data Analysis: A Bayesian Tutorial, Oxford Science Publications (2006)], and a combined Variational/Markov Chain Monte Carlo approach in order to determine the correct decay model. The advantages/shortcomings of these methods are discussed in the context of the decay model selection and parameter estimation using experimentally measured data from real coupled spaces.