--This paper explores the outcome of training state-of-the-art dereverberation models with supervision settings ranging from weakly-supervised to fully unsupervised, relying solely on reverberant signals and an acoustic model for training. Most of the existing deep learning approaches typically require paired dry and reverberant data, which are difficult to obtain in practice. We develop instead a sequential learning strategy motivated by a bayesian formulation of the dereverberation problem, wherein acoustic parameters and dry signals are estimated from reverberant inputs using deep neural networks, guided by a reverberation matching loss. COUSTIC waves propagation in enclosed environments is significantly influenced by reflections and diffractions from surrounding surfaces and objects. These interactions alter the original waveform and result in reverberation, which can be modeled as a superposition of delayed and attenuated versions of the source signal. Reverberation has long been recognized as a critical factor affecting speech intelligibility [1], and its detrimental effects on audio clarity have motivated decades of research. The task of reverberation suppression, commonly referred to as dereverberation, has received renewed attention in recent years due to its relevance in a wide range of audio processing applications. Effective dereverberation is essential in enhancing the performance of hearing aids [2], improving communication quality in hands-free [3] telephony, and enabling robust Automatic Speech Recognition (ASR) in human-machine interaction scenarios [4]. It also serves as a key preprocessing step in general-purpose speech enhancement frameworks [5]. Beyond suppression, reverberation itself plays a constructive role in audio production, particularly in simulating desired acoustic characteristics in post-processing. Reverberation conversion, or acoustic transfer, aims to transform a given recording, possibly containing unknown or undesired room effects, into a version consistent with a target acoustic environment. This work was funded by the European Union (ERC, HI-Audio, 101052978). Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union or the European Research Council.

Estimating frequency-varying acoustic parameters is essential for enhancing immersive perception in realistic spatial audio creation. In this paper, we propose a unified framework that blindly estimates reverberation time (T60), direct-to-reverberant ratio (DRR), and clarity (C50) across 10 frequency bands using first-order Ambisonics (FOA) speech recordings as inputs. The proposed framework utilizes a novel feature named Spectro-Spatial Covariance Vector (SSCV), efficiently representing temporal, spectral as well as spatial information of the FOA signal. Our models significantly outperform existing single-channel methods with only spectral information, reducing estimation errors by more than half for all three acoustic parameters. Additionally, we introduce FOA-Conv3D, a novel back-end network for effectively utilising the SSCV feature with a 3D convolutional encoder. FOA-Conv3D outperforms the convolutional neural network (CNN) and recurrent convolutional neural network (CRNN) backends, achieving lower estimation errors and accounting for a higher proportion of variance (PoV) for all 3 acoustic parameters.

Collaborative robots must effectively communicate their internal state to humans to enable a smooth interaction. Nonverbal communication is widely used to communicate information during human-robot interaction, however, such methods may also be misunderstood, leading to communication errors. In this work, we explore modulating the acoustic parameter values (pitch bend, beats per minute, beats per loop) of nonverbal auditory expressions to convey functional robot states (accomplished, progressing, stuck). We propose a reinforcement learning (RL) algorithm based on noisy human feedback to produce accurately interpreted nonverbal auditory expressions. The proposed approach was evaluated through a user study with 24 participants. The results demonstrate that: 1. Our proposed RL-based approach is able to learn suitable acoustic parameter values which improve the users' ability to correctly identify the state of the robot. 2. Algorithm initialization informed by previous user data can be used to significantly speed up the learning process. 3. The method used for algorithm initialization strongly influences whether participants converge to similar sounds for each robot state. 4. Modulation of pitch bend has the largest influence on user association between sounds and robotic states.

We present a speech database and a phoneme-level language model of Polish. The database and model are designed for the analysis of prosodic and discourse factors and their impact on acoustic parameters in interaction with predictability effects. The database is also the first large, publicly available Polish speech corpus of excellent acoustic quality that can be used for phonetic analysis and training of multi-speaker speech technology systems. The speech in the database is processed in a pipeline that achieves a 90% degree of automation. It incorporates state-of-the-art, freely available tools enabling database expansion or adaptation to additional languages.

We study speech enhancement using deep learning (DL) for virtual meetings on cellular devices, where transmitted speech has background noise and transmission loss that affects speech quality. Since the Deep Noise Suppression (DNS) Challenge dataset of Interspeech 2020 does not contain practical disturbance, we collect a transmitted DNS (t-DNS) dataset using Zoom Meetings over T-Mobile network. We select two baseline models: Demucs and FullSubNet. The Demucs is an endto-end model that takes time-domain inputs and outputs time-domain denoised speech, and the FullSubNet takes time-frequency-domain inputs and outputs the energy ratio of the target speech in the inputs. The goal of this project is to enhance the speech transmitted over the cellular networks using deep learning models.

Despite rapid advancement in recent years, current speech enhancement models often produce speech that differs in perceptual quality from real clean speech. We propose a learning objective that formalizes differences in perceptual quality, by using domain knowledge of acoustic-phonetics. We identify temporal acoustic parameters -- such as spectral tilt, spectral flux, shimmer, etc. -- that are non-differentiable, and we develop a neural network estimator that can accurately predict their time-series values across an utterance. We also model phoneme-specific weights for each feature, as the acoustic parameters are known to show different behavior in different phonemes. We can add this criterion as an auxiliary loss to any model that produces speech, to optimize speech outputs to match the values of clean speech in these features. Experimentally we show that it improves speech enhancement workflows in both time-domain and time-frequency domain, as measured by standard evaluation metrics. We also provide an analysis of phoneme-dependent improvement on acoustic parameters, demonstrating the additional interpretability that our method provides. This analysis can suggest which features are currently the bottleneck for improvement.

Speech enhancement models have greatly progressed in recent years, but still show limits in perceptual quality of their speech outputs. We propose an objective for perceptual quality based on temporal acoustic parameters. These are fundamental speech features that play an essential role in various applications, including speaker recognition and paralinguistic analysis. We provide a differentiable estimator for four categories of low-level acoustic descriptors involving: frequency-related parameters, energy or amplitude-related parameters, spectral balance parameters, and temporal features. Unlike prior work that looks at aggregated acoustic parameters or a few categories of acoustic parameters, our temporal acoustic parameter (TAP) loss enables auxiliary optimization and improvement of many fine-grain speech characteristics in enhancement workflows. We show that adding TAPLoss as an auxiliary objective in speech enhancement produces speech with improved perceptual quality and intelligibility. We use data from the Deep Noise Suppression 2020 Challenge to demonstrate that both time-domain models and time-frequency domain models can benefit from our method.

The primary objective of speech enhancement is to reduce background noise while preserving the target's speech. A common dilemma occurs when a speaker is confined to a noisy environment and receives a call with high background and transmission noise. To address this problem, the Deep Noise Suppression (DNS) Challenge focuses on removing the background noise with the next-generation deep learning models to enhance the target's speech; however, researchers fail to consider Voice Over IP (VoIP) applications their transmission noise. Focusing on Google Meet and its cellular application, our work achieves state-of-the-art performance on the Google Meet To Phone Track of the VoIP DNS Challenge. This paper demonstrates how to beat industrial performance and achieve 1.92 PESQ and 0.88 STOI, as well as superior acoustic fidelity, perceptual quality, and intelligibility in various metrics.

WaveCycleGAN has recently been proposed to bridge the gap between natural and synthesized speech waveforms in statistical parametric speech synthesis and provides fast inference with a moving average model rather than an autoregressive model and high-quality speech synthesis with the adversarial training. However, the human ear can still distinguish the processed speech waveforms from natural ones. One possible cause of this distinguishability is the aliasing observed in the processed speech waveform via down/up-sampling modules. To solve the aliasing and provide higher quality speech synthesis, we propose WaveCycleGAN2, which 1) uses generators without down/up-sampling modules and 2) combines discriminators of the waveform domain and acoustic parameter domain. The results show that the proposed method 1) alleviates the aliasing well, 2) is useful for both speech waveforms generated by analysis-and-synthesis and statistical parametric speech synthesis, and 3) achieves a mean opinion score comparable to those of natural speech and speech synthesized by WaveNet (open WaveNet) and WaveGlow while processing speech samples at a rate of more than 150 kHz on an NVIDIA Tesla P100.