This paper describes Asteroid, the PyTorch-based audio source separation toolkit for researchers. Inspired by the most successful neural source separation systems, it provides all neural building blocks required to build such a system. To improve reproducibility, Kaldi-style recipes on common audio source separation datasets are also provided. This paper describes the software architecture of Asteroid and its most important features. By showing experimental results obtained with Asteroid's recipes, we show that our implementations are at least on par with most results reported in reference papers. The toolkit is publicly available at https://github.com/mpariente/asteroid .

By building upon the recent theory that established the connection between implicit generative modeling (IGM) and optimal transport, in this study, we propose a novel parameter-free algorithm for learning the underlying distributions of complicated datasets and sampling from them. The proposed algorithm is based on a functional optimization problem, which aims at finding a measure that is close to the data distribution as much as possible and also expressive enough for generative modeling purposes. We formulate the problem as a gradient flow in the space of probability measures. The connections between gradient flows and stochastic differential equations let us develop a computationally efficient algorithm for solving the optimization problem. We provide formal theoretical analysis where we prove finite-time error guarantees for the proposed algorithm. To the best of our knowledge, the proposed algorithm is the first nonparametric IGM algorithm with explicit theoretical guarantees. Our experimental results support our theory and show that our algorithm is able to successfully capture the structure of different types of data distributions.

Estimating the maximum number of concurrent speakers from single-channel mixtures is a challenging problem and an essential first step to address various audio-based tasks such as blind source separation, speaker diarization, and audio surveillance. We propose a unifying probabilistic paradigm, where deep neural network architectures are used to infer output posterior distributions. These probabilities are in turn processed to yield discrete point estimates. Designing such architectures often involves two important and complementary aspects that we investigate and discuss. First, we study how recent advances in deep architectures may be exploited for the task of speaker count estimation. In particular, we show that convolutional recurrent neural networks outperform recurrent networks used in a previous study when adequate input features are used. Even for short segments of speech mixtures, we can estimate up to five speakers, with a significantly lower error than other methods. Second, through comprehensive evaluation, we compare the best-performing method to several baselines, as well as the influence of gain variations, different data sets, and reverberation. The output of our proposed method is compared to human performance. Finally, we give insights into the strategy used by our proposed method.

The task of estimating the maximum number of concurrent speakers from single channel mixtures is important for various audio-based applications, such as blind source separation, speaker diarisation, audio surveillance or auditory scene classification. Building upon powerful machine learning methodology, we develop a Deep Neural Network (DNN) that estimates a speaker count. While DNNs efficiently map input representations to output targets, it remains unclear how to best handle the network output to infer integer source count estimates, as a discrete count estimate can either be tackled as a regression or a classification problem. In this paper, we investigate this important design decision and also address complementary parameter choices such as the input representation. We evaluate a state-of-the-art DNN audio model based on a Bi-directional Long Short-Term Memory network architecture for speaker count estimations. Through experimental evaluations aimed at identifying the best overall strategy for the task and show results for five seconds speech segments in mixtures of up to ten speakers.

Dereverberation is useful in hands-free communication and voice controlled devices for distant speech acquisition. Single-channel dereverberation can be achieved by applying a time-frequency (TF) mask to the short-time Fourier transform (STFT) representation of a reverberant signal. Recent approaches have used deep neural networks (DNNs) to estimate such masks. Previously proposed DNN-based mask estimation methods train a DNN to minimize the mean-squared-error (MSE) between the desired and estimated masks. Recent TF mask estimation methods for signal separation directly minimize instead the MSE between the desired and estimated STFT magnitudes. We apply this direct optimization concept to dereverberation. Moreover, as reverberation exceeds the duration of a single STFT frame, we propose to use a bidirectional long short-term memory (LSTM) network which is able to take the relation between multiple STFT frames into account. We evaluated our method for different reverberation times and source-microphone distances using simulated as well as measured room impulse responses of different rooms. An evaluation of the proposed method and a comparison with a state-of-the-art method demonstrate the superiority of our approach and its robustness to different acoustic conditions.

Popular music is often composed of an accompaniment and a lead component, the latter typically consisting of vocals. Filtering such mixtures to extract one or both components has many applications, such as automatic karaoke and remixing. This particular case of source separation yields very specific challenges and opportunities, including the particular complexity of musical structures, but also relevant prior knowledge coming from acoustics, musicology or sound engineering. Due to both its importance in applications and its challenging difficulty, lead and accompaniment separation has been a popular topic in signal processing for decades. In this article, we provide a comprehensive review of this research topic, organizing the different approaches according to whether they are model-based or data-centered. For model-based methods, we organize them according to whether they concentrate on the lead signal, the accompaniment, or both. For data-centered approaches, we discuss the particular difficulty of obtaining data for learning lead separation systems, and then review recent approaches, notably those based on deep learning. Finally, we discuss the delicate problem of evaluating the quality of music separation through adequate metrics and present the results of the largest evaluation, to-date, of lead and accompaniment separation systems. In conjunction with the above, a comprehensive list of references is provided, along with relevant pointers to available implementations and repositories.

In this paper we present a novel source separation method aiming to overcome the difficulty of modelling non-stationary signals. The method can be applied to mixtures of musical instruments with frequency and/or amplitude modulation, e.g. typically caused by vibrato. It is based on a signal representation that divides the complex spectrogram into a grid of patches of arbitrary size. These complex patches are then processed by a two-dimensional discrete Fourier transform, forming a tensor representation which reveals spectral and temporal modulation textures. Our representation can be seen as an alternative to modulation transforms computed on magnitude spectrograms. An adapted factorization model allows to decompose different time-varying harmonic sources based on their particular common modulation profile: hence the name Common Fate Model. The method is evaluated on musical instrument mixtures playing the same fundamental frequency (unison), showing improvement over other state-of-the-art methods.

Estimating the fundamental frequency (F0) of a signal is a well studied task in audio signal processing with many applications. If the F0 varies over time, the complexity increases, and it is also more difficult to provide ground truth data for evaluation. In this paper we present a novel dataset of cello recordings addressing the lack of reference annotations for musical instruments. Besides audio data, we include sensor recordings capturing the finger position on the fingerboard which is converted into an instantaneous frequency estimate. In speech processing, the electroglottograph (EGG) is able to capture the excitation signal of the vocal tract, which is then used to generate a reference instantaneous F0. Inspired by this approach, we included high speed video camera recordings to extract the excitation signal originating from the moving string. The derived data can be used to analyze vibratos --- a very commonly used playing style. The dataset is released under a Creative Commons license.