Low Bit-Rate Wideband Speech Coding: A Deep Generative Model based Approach

TL;DR

This paper introduces a deep generative model-based speech coding method that uses vector quantization of MFCCs and WaveGlow, achieving high-quality wideband speech at low bit-rates with improved efficiency and quality over traditional codecs.

Contribution

It presents a novel deep learning-based speech codec combining VQ of MFCCs and WaveGlow, enabling scalable, high-quality wideband speech coding at low bit-rates.

Findings

Outperforms classic MELPe codec at lower bit-rates

Provides better speech quality on TIMIT corpus

Operates efficiently without autoregressive sampling

Abstract

Traditional low bit-rate speech coding approach only handles narrowband speech at 8kHz, which limits further improvements in speech quality. Motivated by recent successful exploration of deep learning methods for image and speech compression, this paper presents a new approach through vector quantization (VQ) of mel-frequency cepstral coefficients (MFCCs) and using a deep generative model called WaveGlow to provide efficient and high-quality speech coding. The coding feature is sorely an 80-dimension MFCCs vector for 16kHz wideband speech, then speech coding at the bit-rate throughout 1000-2000 bit/s could be scalably implemented by applying different VQ schemes for MFCCs vector. This new deep generative network based codec works fast as the WaveGlow model abandons the sample-by-sample autoregressive mechanism. We evaluated this new approach over the multi-speaker TIMIT corpus, and experimental results demonstrate that it provides better speech quality compared with the state-of-the-art classic MELPe codec at lower bit-rate.