Human perception of the complex world relies on a comprehensive analysis of multi-modal signals, and the co-occurrences of audio and video signals provide humans with rich cues. This paper focuses on novel audio-visual scene synthesis in the real world. Given a video recording of an audio-visual scene, the task is to synthesize new videos with spatial audios along arbitrary novel camera trajectories in that audio-visual scene. Directly using a NeRF-based model for audio synthesis is insufficient due to its lack of prior knowledge and acoustic supervision. To tackle the challenges, we first propose an acoustic-aware audio generation module that integrates our prior knowledge of audio propagation into NeRF, in which we associate audio generation with the 3D geometry of the visual environment. In addition, we propose a coordinate transformation module that expresses a viewing direction relative to the sound source. Such a direction transformation helps the model learn sound source-centric acoustic fields. Moreover, we utilize a head-related impulse response function to synthesize pseudo binaural audio for data augmentation that strengthens training. We qualitatively and quantitatively demonstrate the advantage of our model on real-world audio-visual scenes. We refer interested readers to view our video results for convincing comparisons.
Despite multiple efforts made towards adopting complex-valued deep neural networks (DNNs), it remains an open question whether complex-valued DNNs are generally more effective than real-valued DNNs for monaural speech enhancement. This work is devoted to presenting a critical assessment by systematically examining complex-valued DNNs against their real-valued counterparts. Specifically, we investigate complex-valued DNN atomic units, including linear layers, convolutional layers, long short-term memory (LSTM), and gated linear units. By comparing complex- and real-valued versions of fundamental building blocks in the recently developed gated convolutional recurrent network (GCRN), we show how different mechanisms for basic blocks affect the performance. We also find that the use of complex-valued operations hinders the model capacity when the model size is small. In addition, we examine two recent complex-valued DNNs, i.e. deep complex convolutional recurrent network (DCCRN) and deep complex U-Net (DCUNET). Evaluation results show that both DNNs produce identical performance to their real-valued counterparts while requiring much more computation. Based on these comprehensive comparisons, we conclude that complex-valued DNNs do not provide a performance gain over their real-valued counterparts for monaural speech enhancement, and thus are less desirable due to their higher computational costs.
Audio-visual speech enhancement aims to extract clean speech from a noisy environment by leveraging not only the audio itself but also the target speaker's lip movements. This approach has been shown to yield improvements over audio-only speech enhancement, particularly for the removal of interfering speech. Despite recent advances in speech synthesis, most audio-visual approaches continue to use spectral mapping/masking to reproduce the clean audio, often resulting in visual backbones added to existing speech enhancement architectures. In this work, we propose LA-VocE, a new two-stage approach that predicts mel-spectrograms from noisy audio-visual speech via a transformer-based architecture, and then converts them into waveform audio using a neural vocoder (HiFi-GAN). We train and evaluate our framework on thousands of speakers and 11+ different languages, and study our model's ability to adapt to different levels of background noise and speech interference. Our experiments show that LA-VocE outperforms existing methods according to multiple metrics, particularly under very noisy scenarios.
Most speech enhancement (SE) models learn a point estimate, and do not make use of uncertainty estimation in the learning process. In this paper, we show that modeling heteroscedastic uncertainty by minimizing a multivariate Gaussian negative log-likelihood (NLL) improves SE performance at no extra cost. During training, our approach augments a model learning complex spectral mapping with a temporary submodel to predict the covariance of the enhancement error at each time-frequency bin. Due to unrestricted heteroscedastic uncertainty, the covariance introduces an undersampling effect, detrimental to SE performance. To mitigate undersampling, our approach inflates the uncertainty lower bound and weights each loss component with their uncertainty, effectively compensating severely undersampled components with more penalties. Our multivariate setting reveals common covariance assumptions such as scalar and diagonal matrices. By weakening these assumptions, we show that the NLL achieves superior performance compared to popular losses including the mean squared error (MSE), mean absolute error (MAE), and scale-invariant signal-to-distortion ratio (SI-SDR).
While deep learning based speech enhancement systems have made rapid progress in improving the quality of speech signals, they can still produce outputs that contain artifacts and can sound unnatural. We propose a novel approach to speech enhancement aimed at improving perceptual quality and naturalness of enhanced signals by optimizing for key characteristics of speech. We first identify key acoustic parameters that have been found to correlate well with voice quality (e.g. jitter, shimmer, and spectral flux) and then propose objective functions which are aimed at reducing the difference between clean speech and enhanced speech with respect to these features. The full set of acoustic features is the extended Geneva Acoustic Parameter Set (eGeMAPS), which includes 25 different attributes associated with perception of speech. Given the non-differentiable nature of these feature computation, we first build differentiable estimators of the eGeMAPS and then use them to fine-tune existing speech enhancement systems. Our approach is generic and can be applied to any existing deep learning based enhancement systems to further improve the enhanced speech signals. Experimental results conducted on the Deep Noise Suppression (DNS) Challenge dataset shows that our approach can improve the state-of-the-art deep learning based enhancement systems.
Audio quality assessment is critical for assessing the perceptual realism of sounds. However, the time and expense of obtaining ''gold standard'' human judgments limit the availability of such data. For AR&VR, good perceived sound quality and localizability of sources are among the key elements to ensure complete immersion of the user. Our work introduces SAQAM which uses a multi-task learning framework to assess listening quality (LQ) and spatialization quality (SQ) between any given pair of binaural signals without using any subjective data. We model LQ by training on a simulated dataset of triplet human judgments, and SQ by utilizing activation-level distances from networks trained for direction of arrival (DOA) estimation. We show that SAQAM correlates well with human responses across four diverse datasets. Since it is a deep network, the metric is differentiable, making it suitable as a loss function for other tasks. For example, simply replacing an existing loss with our metric yields improvement in a speech-enhancement network.
Human judgments obtained through Mean Opinion Scores (MOS) are the most reliable way to assess the quality of speech signals. However, several recent attempts to automatically estimate MOS using deep learning approaches lack robustness and generalization capabilities, limiting their use in real-world applications. In this work, we present a novel framework, NORESQA-MOS, for estimating the MOS of a speech signal. Unlike prior works, our approach uses non-matching references as a form of conditioning to ground the MOS estimation by neural networks. We show that NORESQA-MOS provides better generalization and more robust MOS estimation than previous state-of-the-art methods such as DNSMOS and NISQA, even though we use a smaller training set. Moreover, we also show that our generic framework can be combined with other learning methods such as self-supervised learning and can further supplement the benefits from these methods.
We present RemixIT, a simple yet effective self-supervised method for training speech enhancement without the need of a single isolated in-domain speech nor a noise waveform. Our approach overcomes limitations of previous methods which make them dependent on clean in-domain target signals and thus, sensitive to any domain mismatch between train and test samples. RemixIT is based on a continuous self-training scheme in which a pre-trained teacher model on out-of-domain data infers estimated pseudo-target signals for in-domain mixtures. Then, by permuting the estimated clean and noise signals and remixing them together, we generate a new set of bootstrapped mixtures and corresponding pseudo-targets which are used to train the student network. Vice-versa, the teacher periodically refines its estimates using the updated parameters of the latest student models. Experimental results on multiple speech enhancement datasets and tasks not only show the superiority of our method over prior approaches but also showcase that RemixIT can be combined with any separation model as well as be applied towards any semi-supervised and unsupervised domain adaptation task. Our analysis, paired with empirical evidence, sheds light on the inside functioning of our self-training scheme wherein the student model keeps obtaining better performance while observing severely degraded pseudo-targets.