Practical cognitive speech compression

TL;DR

This paper introduces a practical neural speech compression method that achieves low bitrate, low latency, and high subjective quality comparable to standard codecs, suitable for mobile devices.

Contribution

It combines a hierarchical unsupervised encoder with a GAN-based decoder, improving speech quality at low bitrate and demonstrating robustness to quantization.

Findings

Outperforms AMR-WB codec in delay, bitrate, and subjective quality

Robust to quantization of representation features

Suitable for mobile device implementation

Abstract

This paper presents a new neural speech compression method that is practical in the sense that it operates at low bitrate, introduces a low latency, is compatible in computational complexity with current mobile devices, and provides a subjective quality that is comparable to that of standard mobile-telephony codecs. Other recently proposed neural vocoders also have the ability to operate at low bitrate. However, they do not produce the same level of subjective quality as standard codecs. On the other hand, standard codecs rely on objective and short-term metrics such as the segmental signal-to-noise ratio that correlate only weakly with perception. Furthermore, standard codecs are less efficient than unsupervised neural networks at capturing speech attributes, especially long-term ones. The proposed method combines a cognitive-coding encoder that extracts an interpretable unsupervised hierarchical representation with a multi stage decoder that has a GAN-based architecture. We observe that this method is very robust to the quantization of representation features. An AB test was conducted on a subset of the Harvard sentences that are commonly used to evaluate standard mobile-telephony codecs. The results show that the proposed method outperforms the standard AMR-WB codec in terms of delay, bitrate and subjective quality.