Phase-Aware Deep Speech Enhancement: It's All About The Frame Length

TL;DR

This paper investigates how phase and magnitude information contribute to low-latency speech enhancement using DNNs, showing that phase estimation becomes more effective with shorter frames, enabling high-quality, low-latency processing.

Contribution

It systematically studies the role of phase and magnitude in DNN-based speech enhancement across different frame lengths, demonstrating effective phase estimation with short frames.

Findings

DNNs can successfully estimate phase with short frames

Short frames enable low-latency speech enhancement with high quality

Phase-aware DNNs outperform magnitude-only approaches at low latency

Abstract

Algorithmic latency in speech processing is dominated by the frame length used for Fourier analysis, which in turn limits the achievable performance of magnitude-centric approaches. As previous studies suggest the importance of phase grows with decreasing frame length, this work presents a systematical study on the contribution of phase and magnitude in modern Deep Neural Network (DNN)-based speech enhancement at different frame lengths. Results indicate that DNNs can successfully estimate phase when using short frames, with similar or better overall performance compared to using longer frames. Thus, interestingly, modern phase-aware DNNs allow for low-latency speech enhancement at high quality.