Contrastive Representation Learning for Acoustic Parameter Estimation

TL;DR

This paper introduces a contrastive learning method to derive low-dimensional acoustic environment representations from reverberant speech, enabling effective parameter estimation and room classification with generalization capabilities.

Contribution

It presents a novel contrastive learning approach utilizing RIR convolution for data augmentation, improving acoustic parameter estimation and room classification from single-channel speech.

Findings

Embeddings perform well on unseen data

Comparable to fully-supervised methods

Effective across multiple downstream tasks

Abstract

A study is presented in which a contrastive learning approach is used to extract low-dimensional representations of the acoustic environment from single-channel, reverberant speech signals. Convolution of room impulse responses (RIRs) with anechoic source signals is leveraged as a data augmentation technique that offers considerable flexibility in the design of the upstream task. We evaluate the embeddings across three different downstream tasks, which include the regression of acoustic parameters reverberation time RT60 and clarity index C50, and the classification into small and large rooms. We demonstrate that the learned representations generalize well to unseen data and perform similarly to a fully-supervised baseline.