Neural Steerer: Novel Steering Vector Synthesis with a Causal Neural Field over Frequency and Source Positions

TL;DR

This paper introduces Neural Steerer, a neural field model that accurately interpolates steering vectors across frequency and source positions, reducing measurement resources for sound localization and separation.

Contribution

The paper presents a novel causal neural field model for continuous steering vector synthesis, incorporating phase difference and causality constraints, inspired by neural fields in computer vision.

Findings

Effective interpolation of measured steering vectors demonstrated

Reduces measurement resources needed for sound source localization

Outperforms classical linear interpolation methods

Abstract

We address the problem of accurately interpolating measured anechoic steering vectors with a deep learning framework called the neural field. This task plays a pivotal role in reducing the resource-intensive measurements required for precise sound source separation and localization, essential as the front-end of speech recognition. Classical approaches to interpolation rely on linear weighting of nearby measurements in space on a fixed, discrete set of frequencies. Drawing inspiration from the success of neural fields for novel view synthesis in computer vision, we introduce the neural steerer, a continuous complex-valued function that takes both frequency and direction as input and produces the corresponding steering vector. Importantly, it incorporates inter-channel phase difference information and a regularization term enforcing filter causality, essential for accurate steering vector modeling. Our experiments, conducted using a dataset of real measured steering vectors, demonstrate the effectiveness of our resolution-free model in interpolating such measurements.