A Low-Complexity Speech Codec Using Parametric Dithering for ASR

TL;DR

This paper introduces a low-complexity speech codec using parametric dithering that enhances ASR performance at low bit resolutions, achieving significant CER improvements and adaptable data rate control.

Contribution

It proposes a novel parametric dithering technique for lossy speech compression, improving ASR accuracy at low bit resolutions with a flexible, low-complexity pipeline.

Findings

25% CER reduction at 1-bit resolution

Over 32% CER improvement at 2- and 3-bit resolutions

Codec adaptability to performance and entropy constraints

Abstract

Dithering is a technique commonly used to improve the perceptual quality of lossy data compression. In this work, we analytically and experimentally justify the use of dithering for ASR input compression. We formalize an understanding of optimal ASR performance under lossy input compression and leverage this to propose a parametric dithering technique for a low-complexity speech compression pipeline. The method performs well at 1-bit resolution, showing a 25\% relative CER improvement, while also demonstrating improvements of 32.4\% and 33.5\% at 2- and 3-bit resolution, respectively, with our second dither choice yielding a reduced data rate. The proposed codec is adaptable to meet performance targets or stay within entropy constraints.