Self-Supervised Learning of Spatial Acoustic Representation with Cross-Channel Signal Reconstruction and Multi-Channel Conformer

TL;DR

This work tackles the simulation-to-reality generalization gap in spatial acoustic parameter estimation by learning a universal spatial representation from unlabeled dual-channel recordings through self-supervision. It introduces a cross-channel signal reconstruction pretext task to disentangle spatial information from signal content, and employs a Multi-Channel Conformer encoder to capture local and global time-frequency dependencies. The approach is pre-trained on unlabeled data and fine-tuned on small labeled sets for downstream tasks such as TDOA, DRR, $T_{60}$, $C_{50}$, and absorption estimates, demonstrating improvements over fully supervised learning in simulated and real data. The results indicate strong potential for real-world deployment where annotations are scarce, and suggest broader applicability to multi-channel audio processing. Overall, this is the first reported self-supervised method that learns spatial acoustic representations tailored to multi-channel audio and downstream spatial parameter estimation.

Abstract

Supervised learning methods have shown effectiveness in estimating spatial acoustic parameters such as time difference of arrival, direct-to-reverberant ratio and reverberation time. However, they still suffer from the simulation-to-reality generalization problem due to the mismatch between simulated and real-world acoustic characteristics and the deficiency of annotated real-world data. To this end, this work proposes a self-supervised method that takes full advantage of unlabeled data for spatial acoustic parameter estimation. First, a new pretext task, i.e. cross-channel signal reconstruction (CCSR), is designed to learn a universal spatial acoustic representation from unlabeled multi-channel microphone signals. We mask partial signals of one channel and ask the model to reconstruct them, which makes it possible to learn spatial acoustic information from unmasked signals and extract source information from the other microphone channel. An encoder-decoder structure is used to disentangle the two kinds of information. By fine-tuning the pre-trained spatial encoder with a small annotated dataset, this encoder can be used to estimate spatial acoustic parameters. Second, a novel multi-channel audio Conformer (MC-Conformer) is adopted as the encoder model architecture, which is suitable for both the pretext and downstream tasks. It is carefully designed to be able to capture the local and global characteristics of spatial acoustics exhibited in the time-frequency domain. Experimental results of five acoustic parameter estimation tasks on both simulated and real-world data show the effectiveness of the proposed method. To the best of our knowledge, this is the first self-supervised learning method in the field of spatial acoustic representation learning and multi-channel audio signal processing.

Figures (9)

• Figure 1: Illustration of self-supervised learning of spatial acoustic representation using multi-channel microphone recordings. The direct path, early reflections and late reverberation are illustrated in red, green and blue colors, respectively.
• Figure 2: Block diagram of the proposed self-supervised spatial acoustic representation learning model. The complex-valued STFT coefficients are illustrated by their real-part spectrograms.
• Figure 3: Model architecture of (a) spatial/spectral encoder (namely multi-channel audio Conformer), (b) decoder and (c) convolution block in the encoder. $D$ is the hidden dimension, and $D=D^{\rm{spat}}$ for spatial encoder and $D=D^{\rm{spec}}$ for spectral encoder.
• Figure 4: Results of TDOA, DRR, $T_{60}$, $C_{50}$ and absorption coefficient (ABS) estimation on the simulated dataset, for the proposed self-supervised pre-training plus fine-tuning method and the fully supervised training method, when using labeled data from different amounts of training rooms.
• Figure 5: Learning curves (MAE versus training iteration) for TDOA, DRR and $T_{60}$ estimation on the simulated dataset, for the proposed fine-tuning scheme and the scheme of training from scratch. The number of training rooms is 8.