In recent years, substantial progress has been made in the field of reverberant speech signal processing, including both single- and multichannel dereverberation techniques and automatic speech recognition (ASR) techniques that are robust to reverberation. In this paper, we describe the REVERB challenge, which is an evaluation campaign that was designed to evaluate such speech enhancement (SE) and ASR techniques to reveal the state-of-the-art techniques and obtain new insights regarding potential future research directions. Even though most existing benchmark tasks and challenges for distant speech processing focus on the noise robustness issue and sometimes only on a single-channel scenario, a particular novelty of the REVERB challenge is that it is carefully designed to test robustness against reverberation, based on both real, single-channel, and multichannel recordings. This challenge attracted 27 papers, which represent 25 systems specifically designed for SE purposes and 49 systems specifically designed for ASR purposes. This paper describes the problems dealt within the challenge, provides an overview of the submitted systems, and scrutinizes them to clarify what current processing strategies appear effective in reverberant speech processing.

Supervised learning based methods for source localization, being data driven, can be adapted to different acoustic conditions via training and have been shown to be robust to adverse acoustic environments. In this paper, a convolutional neural network (CNN) based supervised learning method for estimating the direction-of-arrival (DOA) of multiple speakers is proposed. Multi-speaker DOA estimation is formulated as a multi-class multi-label classification problem, where the assignment of each DOA label to the input feature is treated as a separate binary classification problem. The phase component of the short-time Fourier transform (STFT) coefficients of the received microphone signals are directly fed into the CNN, and the features for DOA estimation are learnt during training. Utilizing the assumption of disjoint speaker activity in the STFT domain, a novel method is proposed to train the CNN with synthesized noise signals. Through experimental evaluation with both simulated and measured acoustic impulse responses, the ability of the proposed DOA estimation approach to adapt to unseen acoustic conditions and its robustness to unseen noise type is demonstrated. Through additional empirical investigation, it is also shown that with an array of M microphone our proposed framework yields the best localization performance with M-1 convolution layers. The ability of the proposed method to accurately localize speakers in a dynamic acoustic scenario with varying number of sources is also shown.

Recently, substantial progress has been made in the field of reverberant speech signal processing, including both single- and multichannel dereverberation techniques, and automatic speech recognition (ASR) techniques robust to reverberation. To evaluate state-of-the-art algorithms and obtain new insights regarding potential future research directions, we propose a common evaluation framework including datasets, tasks, and evaluation metrics for both speech enhancement and ASR techniques. The proposed framework will be used as a common basis for the REVERB (REverberant Voice Enhancement and Recognition Benchmark) challenge. This paper describes the rationale behind the challenge, and provides a detailed description of the evaluation framework and benchmark results.

A convolution neural network (CNN) based classification method for broadband DOA estimation is proposed, where the phase component of the short-time Fourier transform coefficients of the received microphone signals are directly fed into the CNN and the features required for DOA estimation are learned during training. Since only the phase component of the input is used, the CNN can be trained with synthesized noise signals, thereby making the preparation of the training data set easier compared to using speech signals. Through experimental evaluation, the ability of the proposed noise trained CNN framework to generalize to speech sources is demonstrated. In addition, the robustness of the system to noise, small perturbations in microphone positions, as well as its ability to adapt to different acoustic conditions is investigated using experiments with simulated and real data.

Researchers in the signal processing community often require sensor signals that result from a spherically or cylindrically isotropic noise field for simulation purposes. Although it has been shown that these signals can be generated using a number of uncorrelated noise sources that are uniformly spaced on a sphere or cylinder, this method is seldom used in practice. In this paper algorithms that generate sensor signals of an arbitrary one- and three-dimensional array that result from a spherically or cylindrically isotropic noise field are developed. Furthermore, the influence of the number of noise sources on the accuracy of the generated sensor signals is investigated.

Noise fields encountered in real-life scenarios can often be approximated as spherical or cylindrical noise fields. The characteristics of the noise field can be described by a spatial coherence function. For simulation purposes, researchers in the signal processing community often require sensor signals that exhibit a specific spatial coherence function. In addition, they often require a specific type of noise such as temporally correlated noise, babble speech that comprises a mixture of mutually independent speech fragments, or factory noise. Existing algorithms are unable to generate sensor signals such as babble speech and factory noise observed in an arbitrary noise field. In this paper an efficient algorithm is developed that generates multisensor signals under a predefined spatial coherence constraint. The benefit of the developed algorithm is twofold. Firstly, there are no restrictions on the spatial coherence function. Secondly, to generate M sensor signals the algorithm requires only M mutually independent noise signals. The performance evaluation shows that the developed algorithm is able to generate a more accurate spatial coherence between the generated sensor signals compared to the so-called image method that is frequently used in the signal processing community.