Evaluation of Data-Driven Room Geometry Inference Methods Using a Smart Speaker Prototype

Recent studies tackling the problem of room geometry inference (RGI) with data-driven methods require a substantial amount of room impulse responses (RIRs) collected in a diverse set of rooms for training the deep neural networks (DNNs). However, this may be a prohibitively time-consuming and labor-intensive task, which requires simulated data. This study explores regularization methods to improve RGI accuracy when DNNs are trained with simulated data and tested with measured data. We use a smart speaker prototype equipped with multiple microphones and directional loudspeakers for real-world RIR measurements. The results indicate that applying dropout at the network’s input layer results in improved generalization compared to using it solely in the hidden layers. Moreover, RGI using multiple directional loudspeakers leads to increased estimation accuracy when compared to the single loudspeaker case, mitigating the impact of source directivity.

1. INTRODUCTION

The sound field inside a room is significantly affected by the shape of the room. Knowledge of room dimensions and layout, embedded in room impulse responses (RIRs), enhances various audio-related tasks [1]. These tasks include sound source tracking [2], dereverberation, [3] room acoustic parameter estimation [4], and auralization [5]. The goal of extracting this information by utilizing RIRs is known as room geometry inference (RGI), and it has been studied extensively [6]-[9].