We present SoundStream, a novel neural audio codec that can efficiently compress speech, music and general audio at bitrates normally targeted by speech-tailored codecs. SoundStream relies on a model architecture composed by a fully convolutional encoder/decoder network and a residual vector quantizer, which are trained jointly end-to-end. Training leverages recent advances in text-to-speech and speech enhancement, which combine adversarial and reconstruction losses to allow the generation of high-quality audio content from quantized embeddings. By training with structured dropout applied to quantizer layers, a single model can operate across variable bitrates from 3kbps to 18kbps, with a negligible quality loss when compared with models trained at fixed bitrates. In addition, the model is amenable to a low latency implementation, which supports streamable inference and runs in real time on a smartphone CPU. In subjective evaluations using audio at 24kHz sampling rate, SoundStream at 3kbps outperforms Opus at 12kbps and approaches EVS at 9.6kbps. Moreover, we are able to perform joint compression and enhancement either at the encoder or at the decoder side with no additional latency, which we demonstrate through background noise suppression for speech.

Streaming services seek to optimise their use of bandwidth across audio and visual channels to maximise the quality of experience for users. This letter evaluates whether objective quality metrics can predict the audio quality for music encoded at low bitrates by comparing objective predictions with results from listener tests. Three objective metrics were benchmarked: PEAQ, POLQA, and VISQOLAudio. The results demonstrate objective metrics designed for speech quality assessment have a strong potential for quality assessment of low bitrate audio codecs.

This letter presents a phase-sensitive joint learning algorithm for single-channel speech enhancement. Although a deep learning framework that estimates time-frequency (T-F) domain ideal ratio masks demonstrates a strong performance, it is limited in that the enhancement process is performed only in the magnitude domain, while the phase spectra remain unchanged. Thus, recent studies have been conducted to involve phase spectra in speech enhancement systems. A phase-sensitive mask (PSM) is a T-F mask that implicitly represents phase-related information. However, since the PSM has an unbounded value, the networks are trained to target its truncated values rather than directly estimating it. To effectively train the PSM, we first approximate it to have a bounded dynamic range under the assumption that speech and noise are uncorrelated. We then propose a joint learning algorithm that trains the approximated value through its parameterized variables in order to minimize the inevitable error caused by the truncation process. Specifically, we design a network that explicitly targets three parameterized variables: speech magnitude spectra, noise magnitude spectra, and phase difference of clean to noisy spectra. To further improve the performance, we also investigate how the dynamic range of magnitude spectra controlled by a warping function affects the final performance in joint learning algorithms. Finally, we examined how the proposed additional constraint that preserves the sum of the estimated speech and noise power spectra affects the overall system performance. The experimental results show that the proposed learning algorithm outperforms the conventional learning algorithm with the truncated phase-sensitive approximation.

Spatial audio is essential for creating a sense of immersion in virtual environments. Efficient encoding methods are required to deliver spatial audio over networks without compromising Quality of Service (QoS). Streaming service providers such as YouTube typically transcode content into various bit rates and need a perceptually relevant audio quality metric to monitor users’ perceived quality and spatial localization accuracy. The aim of the paper is two-fold. First, it is to investigate the effect of Opus codec compression on the quality of spatial audio as perceived by listeners using subjective listening tests. Secondly, it is to introduce AMBIQUAL, a full reference objective metric for spatial audio quality, which derives both listening quality and localization accuracy metrics directly from the B-format Ambisonic audio. We compare AMBIQUAL quality predictions with subjective quality assessments across a variety of audio samples which have been compressed using the Opus 1.2 codec at various bit rates. Listening quality and localization accuracy of first and third-order Ambisonics were evaluated. Several fixed and dynamic audio sources (single and multiple) were used to evaluate localization accuracy. Results show good correlation regarding listening quality and localization accuracy between objective quality scores using AMBIQUAL and subjective scores obtained during listening tests.

The increasing popularity of Ambisonics as a spatial audio format for streaming services poses new challenges to existing audio coding techniques. Immersive audio delivered to mobile devices requires an efficient bitrate compression that does not affect the spatial quality of the content. Good localizability of virtual sound sources is one of the key elements that must be preserved. This study was conducted to investigate the localization precision of virtual sound source presentations within Ambisonic scenes encoded with Opus low-bitrate compression at different bitrates and Ambisonic orders (1st, 3rd, and 5th). The test stimuli were reproduced over a 50-channel spherical loudspeaker configuration and binaurally using individually measured and generic Head-Related Transfer Functions (HRTFs). Participants were asked to adjust the position of a virtual acoustic pointer to match the position of virtual sound source within the bitrate-compressed Ambisonic scene. Results show that auditory localization in low-bitrate compressed Ambisonic scenes is not significantly affected by codec parameters. The key factors influencing localization are the rendering method and Ambisonic order truncation. This suggests that efficient perceptual coding might be successfully used for mobile spatial audio delivery.

This article explores the integration of model-based and data-driven approaches within the realm of neural speech and audio coding systems. It highlights the challenges posed by the subjective evaluation processes of speech and audio codecs and discusses the limitations of purely data-driven approaches, which often require inefficiently large architectures to match the performance of model-based methods. The study presents hybrid systems as a viable solution, offering significant improvements to the performance of conventional codecs through meticulously chosen design enhancements. Specifically, it introduces a neural network-based signal enhancer that is designed to postprocess existing codecs’ output, along with the autoencoder-based end-to-end models and LPCNet–hybrid systems that combine linear predictive coding (LPC) with neural networks. Furthermore, the article delves into predictive models that operate within custom feature spaces (TF-Codec) or predefined transform domains (MDCTNet) and examines the use of psychoacoustically calibrated loss functions to train end-to-end neural audio codecs. Through these investigations, the article demonstrates the potential of hybrid systems to advance the field of speech and audio coding by bridging the gap between traditional model-based approaches and modern data-driven techniques.