Modeling room impulse response via composites of spatial-temporal Gaussian processes

This paper develops a novel spatial-temporal class of semi-parametric statistical models to estimate a transfer function of a time-domain room impulse response for reverberation in closed environments. This novel approach involves developing two non-parameteric models, one for the early phase and the other for the late phase for reverberation. The models are based on a composite of two Gaussian Process(GP) structures. The paper investigates the impact of the choice of spatial and temporal kernels on the estimation and prediction performance. The proposed framework incorporates as a special case the widely utilized exponentially-decaying model and extends this model structure within the GP setting to more advanced spatial-temporal structures suitable to perform estimation of the reverberant transfer function.