Maximum-likelihood based 3D acoustical signature estimation

Abstract

An audio recording, made in a real environment, carries an acoustical signature which changes according to the acoustical characteristics of the environment and the recording positions. This signature which is similar to a 3D room impulse response contains the directions, levels and arrival times of the direct source and reflections. Although it is easy to obtain reverberant recordings by convolving clean recordings with the acoustical signature, estimating the signature from any recording is a difficult inverse problem. Acoustical signature estimation is important in acoustical analysis, audio forensics for authentication, room size and shape estimation and improving speech intelligibility by dereverberation. In this work, the statistical modelling of intensity vector directions, which are obtained from compact microphone array recordings is made. Obtained statistical distribution is used for reducing the reverberation based on the maximum-likelihood estimation method. This dereverberated sound enables deconvolving the reverberant recordings to estimate the acoustical signature.