Hearing in a shoe-box: Binaural source position and wall absorption estimation using virtually supervised learning

Abstract

This paper introduces a new framework for supervised sound source localization referred to as virtually-supervised learning. An acoustic shoe-box room simulator is used to generate a large number of binaural single-source audio scenes. These scenes are used to build a dataset of spatial binaural features annotated with acoustic properties such as the 3D source position and the walls' absorption coefficients. A probabilistic high-to low-dimensional regression framework is used to learn a mapping from these features to the acoustic properties. Results indicate that this mapping successfully estimates the azimuth and elevation of new sources, but also their range and even the walls' absorption coefficients solely based on binaural signals. Results also reveal that incorporating randomdiffusion effects in the data significantly improves the estimation of all parameters.

VAST: The Virtual Acoustic Space Traveler Dataset

“…This model was notably showed to realistically account for sound scattering due to the presence of objects, by comparing simulated RIRs with measured ones in [22]. The study [8] suggests that such diffusion effects play an important role in sound source localization. VAST KEMAR 0 contains over 110, 000 RIR, which required about 700 CPU-hours of computation.…”

SofaMyRoom: a fast and multiplatform "shoebox" room simulator for binaural room impulse response dataset generation

“…Finally, the integration with the VAST toolbox [16] allows the user to easily generate arbitrarily large datasets of BRIRs. The provided MATLAB scripts allow to: (i) initialize an empty VAST structure, (ii) define the room acoustic conditions and (iii) automatically populate the dataset with metadata while calling SofaMyRoom to generate and store BRIRs.…”

“…Data-driven and machine-hearing systems are becoming a key component in audio signal processing research, but they require large amount of labelled data in order to be deployed [16]. On the other hand, the process of recording such data from a real environment involves manual operations that are time consuming and error-prone.…”

Feature Enhancement with Deep Feature Losses for Speaker Verification

“…For the final evaluation, we use standard metrics like Equal Error Rate (EER) and Minimum Decision Cost Function (minDCF) at target prior p = 0.05 (NIST SRE18 VAST operating point). The Code for this work is available online 1 and a parent paper is submitted in parallel [22].…”

“…Various phenomena degrades speech such as noise, reverberation, speaker movement, device orientation, and room characteristics [1]. This makes the deployment of Speaker Verification (SV) systems challenging.…”