The audio source separation tasks, such as speech enhancement, speech separation, and music source separation, have achieved impressive performance in recent studies. The powerful modeling capabilities of deep neural networks give us hope for more challenging tasks. This paper launches a new multi-task audio source separation (MTASS) challenge to separate the speech, music, and noise signals from the monaural mixture. First, we introduce the details of this task and generate a dataset of mixtures containing speech, music, and background noises. Then, we propose an MTASS model in the complex domain to fully utilize the differences in spectral characteristics of the three audio signals. In detail, the proposed model follows a two-stage pipeline, which separates the three types of audio signals and then performs signal compensation separately. After comparing different training targets, the complex ratio mask is selected as a more suitable target for the MTASS. The experimental results also indicate that the residual signal compensation module helps to recover the signals further. The proposed model shows significant advantages in separation performance over several well-known separation models.
This paper describes a hands-on comparison on using state-of-the-art music source separation deep neural networks (DNNs) before and after task-specific fine-tuning for separating speech content from non-speech content in broadcast audio (i.e., dialog separation). The music separation models are selected as they share the number of channels (2) and sampling rate (44.1 kHz or higher) with the considered broadcast content, and vocals separation in music is considered as a parallel for dialog separation in the target application domain. These similarities are assumed to enable transfer learning between the tasks. Three models pre-trained on music (Open-Unmix, Spleeter, and Conv-TasNet) are considered in the experiments, and fine-tuned with real broadcast data. The performance of the models is evaluated before and after fine-tuning with computational evaluation metrics (SI-SIRi, SI-SDRi, 2f-model), as well as with a listening test simulating an application where the non-speech signal is partially attenuated, e.g., for better speech intelligibility. The evaluations include two reference systems specifically developed for dialog separation. The results indicate that pre-trained music source separation models can be used for dialog separation to some degree, and that they benefit from the fine-tuning, reaching a performance close to task-specific solutions.
Audio source separation is often used as preprocessing of various applications, and one of its ultimate goals is to construct a single versatile model capable of dealing with the varieties of audio signals. Since sampling frequency, one of the audio signal varieties, is usually application specific, the preceding audio source separation model should be able to deal with audio signals of all sampling frequencies specified in the target applications. However, conventional models based on deep neural networks (DNNs) are trained only at the sampling frequency specified by the training data, and there are no guarantees that they work with unseen sampling frequencies. In this paper, we propose a convolution layer capable of handling arbitrary sampling frequencies by a single DNN. Through music source separation experiments, we show that the introduction of the proposed layer enables a conventional audio source separation model to consistently work with even unseen sampling frequencies.
The task of isolating a target singing voice in music videos has useful applications. In this work, we explore the single-channel singing voice separation problem from a multimodal perspective, by jointly learning from audio and visual modalities. To do so, we present Acappella, a dataset spanning around 46 hours of a cappella solo singing videos sourced from YouTube. We also propose an audio-visual convolutional network based on graphs which achieves state-of-the-art singing voice separation results on our dataset and compare it against its audio-only counterpart, U-Net, and a state-of-the-art audio-visual speech separation model. We evaluate the models in the following challenging setups: i) presence of overlapping voices in the audio mixtures, ii) the target voice set to lower volume levels in the mix, and iii) combination of i) and ii). The third one being the most challenging evaluation setup. We demonstrate that our model outperforms the baseline models in the singing voice separation task in the most challenging evaluation setup. The code, the pre-trained models, and the dataset are publicly available at https://ipcv.github.io/Acappella/able at https://ipcv.github.io/Acappella/
The advent of deep learning has led to the prevalence of deep neural network architectures for monaural music source separation, with end-to-end approaches that operate directly on the waveform level increasingly receiving research attention. Among these approaches, transformation of the input mixture to a learned latent space, and multiplicative application of a soft mask to the latent mixture, achieves the best performance, but is prone to the introduction of artifacts to the source estimate. To alleviate this problem, in this paper we propose a hybrid time-domain approach, termed the HTMD-Net, combining a lightweight masking component and a denoising module, based on skip connections, in order to refine the source estimated by the masking procedure. Evaluation of our approach in the task of monaural singing voice separation in the musdb18 dataset indicates that our proposed method achieves competitive performance compared to methods based purely on masking when trained under the same conditions, especially regarding the behavior during silent segments, while achieving higher computational efficiency.
Recently, significant progress has been made in audio source separation by the application of deep learning techniques. Current methods that combine both audio and visual information use 2D representations such as images to guide the separation process. However, in order to (re)-create acoustically correct scenes for 3D virtual/augmented reality applications from recordings of real music ensembles, detailed information about each sound source in the 3D environment is required. This demand, together with the proliferation of 3D visual acquisition systems like LiDAR or rgb-depth cameras, stimulates the creation of models that can guide the audio separation using 3D visual information. This paper proposes a multi-modal deep learning model to perform music source separation conditioned on 3D point clouds of music performance recordings. This model extracts visual features using 3D sparse convolutions, while audio features are extracted using dense convolutions. A fusion module combines the extracted features to finally perform the audio source separation. It is shown, that the presented model can distinguish the musical instruments from a single 3D point cloud frame, and perform source separation qualitatively similar to a reference case, where manually assigned instrument labels are provided.
This paper addresses the problem of domain adaptation for the task of music source separation. Using datasets from two different domains, we compare the performance of a deep learning-based harmonic-percussive source separation model under different training scenarios, including supervised joint training using data from both domains and pre-training in one domain with fine-tuning in another. We propose an adversarial unsupervised domain adaptation approach suitable for the case where no labelled data (ground-truth source signals) from a target domain is available. By leveraging unlabelled data (only mixtures) from this domain, experiments show that our framework can improve separation performance on the new domain without losing any considerable performance on the original domain. The paper also introduces the Tap & Fiddle dataset, a dataset containing recordings of Scandinavian fiddle tunes along with isolated tracks for 'foot-tapping' and 'violin'.
Given recent advances in deep music source separation, we propose a feature representation method that combines source separation with a state-of-the-art representation learning technique that is suitably repurposed for computer audition (i.e. machine listening). We train a depthwise separable convolutional neural network on a challenging electronic dance music (EDM) data set and compare its performance to convolutional neural networks operating on both source separated and standard spectrograms. It is shown that source separation improves classification performance in a limited-data setting compared to the standard single spectrogram approach.
This paper proposes several improvements for music separation with deep neural networks (DNNs), namely a multi-domain loss (MDL) and two combination schemes. First, by using MDL we take advantage of the frequency and time domain representation of audio signals. Next, we utilize the relationship among instruments by jointly considering them. We do this on the one hand by modifying the network architecture and introducing a CrossNet structure. On the other hand, we consider combinations of instrument estimates by using a new combination loss (CL). MDL and CL can easily be applied to many existing DNN-based separation methods as they are merely loss functions which are only used during training and which do not affect the inference step. Experimental results show that the performance of Open-Unmix (UMX), a well-known and state-of-the-art open source library for music separation, can be improved by utilizing our above schemes. Our modifications of UMX are open-sourced together with this paper.
Tasks that involve high-resolution dense prediction require a modeling of both local and global patterns in a large input field. Although the local and global structures often depend on each other and their simultaneous modeling is important, many convolutional neural network (CNN)-based approaches interchange representations in different resolutions only a few times. In this paper, we claim the importance of a dense simultaneous modeling of multiresolution representation and propose a novel CNN architecture called densely connected multidilated DenseNet (D3Net). D3Net involves a novel multidilated convolution that has different dilation factors in a single layer to model different resolutions simultaneously. By combining the multidilated convolution with the DenseNet architecture, D3Net incorporates multiresolution learning with an exponentially growing receptive field in almost all layers, while avoiding the aliasing problem that occurs when we naively incorporate the dilated convolution in DenseNet. Experiments on the image semantic segmentation task using Cityscapes and the audio source separation task using MUSDB18 show that the proposed method has superior performance over state-of-the-art methods.
Supervised deep learning methods for performing audio source separation can be very effective in domains where there is a large amount of training data. While some music domains have enough data suitable for training a separation system, such as rock and pop genres, many musical domains do not, such as classical music, choral music, and non-Western music traditions. It is well known that transferring learning from related domains can result in a performance boost for deep learning systems, but it is not always clear how best to do pretraining. In this work we investigate the effectiveness of data augmentation during pretraining, the impact on performance as a result of pretraining and downstream datasets having similar content domains, and also explore how much of a model must be retrained on the final target task, once pretrained.
Recent deep-learning approaches have shown that Frequency Transformation (FT) blocks can significantly improve spectrogram-based single-source separation models by capturing frequency patterns. The goal of this paper is to extend the FT block to fit the multi-source task. We propose the Latent Source Attentive Frequency Transformation (LaSAFT) block to capture source-dependent frequency patterns. We also propose the Gated Point-wise Convolutional Modulation (GPoCM), an extension of Feature-wise Linear Modulation (FiLM), to modulate internal features. By employing these two novel methods, we extend the Conditioned-U-Net (CUNet) for multi-source separation, and the experimental results indicate that our LaSAFT and GPoCM can improve the CUNet's performance, achieving state-of-the-art SDR performance on several MUSDB18 source separation tasks.
Most music source separation systems require large collections of isolated sources for training, which can be difficult to obtain. In this work, we use musical scores, which are comparatively easy to obtain, as a weak label for training a source separation system. In contrast with previous score-informed separation approaches, our system does not require isolated sources, and score is used only as a training target, not required for inference. Our model consists of a separator that outputs a time-frequency mask for each instrument, and a transcriptor that acts as a critic, providing both temporal and frequency supervision to guide the learning of the separator. A harmonic mask constraint is introduced as another way of leveraging score information during training, and we propose two novel adversarial losses for additional fine-tuning of both the transcriptor and the separator. Results demonstrate that using score information outperforms temporal weak-labels, and adversarial structures lead to further improvements in both separation and transcription performance.
Music source separation represents the task of extracting all the instruments from a given song. Recent breakthroughs on this challenge have gravitated around a single dataset, MUSDB, only limited to four instrument classes. Larger datasets and more instruments are costly and time-consuming in collecting data and training deep neural networks (DNNs). In this work, we propose a fast method to evaluate the separability of instruments in any dataset without training and tuning a DNN. This separability measure helps to select appropriate samples for the efficient training of neural networks. Based on the oracle principle with an ideal ratio mask, our approach is an excellent proxy to estimate the separation performances of state-of-the-art deep learning approaches such as TasNet or Open-Unmix. Our results contribute to revealing two essential points for audio source separation: 1) the ideal ratio mask, although light and straightforward, provides an accurate measure of the audio separability performance of recent neural nets, and 2) new end-to-end learning methods such as Tasnet, that operate directly on waveforms, are, in fact, internally building a Time-Frequency (TF) representation, so that they encounter the same limitations as the TF based-methods when separating audio pattern overlapping in the TF plane.
Music source separation involves a large input field to model a long-term dependence of an audio signal. Previous convolutional neural network (CNN)-based approaches address the large input field modeling using sequentially down- and up-sampling feature maps or dilated convolution. In this paper, we claim the importance of a rapid growth of a receptive field and a simultaneous modeling of multi-resolution data in a single convolution layer, and propose a novel CNN architecture called densely connected dilated DenseNet (D3Net). D3Net involves a novel multi-dilated convolution that has different dilation factors in a single layer to model different resolutions simultaneously. By combining the multi-dilated convolution with DenseNet architecture, D3Net avoids the aliasing problem that exists when we naively incorporate the dilated convolution in DenseNet. Experimental results on MUSDB18 dataset show that D3Net achieves state-of-the-art performance with an average signal to distortion ratio (SDR) of 6.01 dB.
Blind music source separation has been a popular and active subject of research in both the music information retrieval and signal processing communities. To counter the lack of available multi-track data for supervised model training, a data augmentation method that creates artificial mixtures by combining tracks from different songs has been shown useful in recent works. Following this light, we examine further in this paper extended data augmentation methods that consider more sophisticated mixing settings employed in the modern music production routine, the relationship between the tracks to be combined, and factors of silence. As a case study, we consider the separation of violin and piano tracks in a violin piano ensemble, evaluating the performance in terms of common metrics, namely SDR, SIR, and SAR. In addition to examining the effectiveness of these new data augmentation methods, we also study the influence of the amount of training data. Our evaluation shows that the proposed mixing-specific data augmentation methods can help improve the performance of a deep learning-based model for source separation, especially in the case of small training data.
Music source separation is a core task in music information retrieval which has seen a dramatic improvement in the past years. Nevertheless, most of the existing systems focus exclusively on the problem of source separation itself and ignore the utilization of other~---possibly related---~MIR tasks which could lead to additional quality gains. In this work, we propose a novel multitask structure to investigate using instrument activation information to improve source separation performance. Furthermore, we investigate our system on six independent instruments, a more realistic scenario than the three instruments included in the widely-used MUSDB dataset, by leveraging a combination of the MedleyDB and Mixing Secrets datasets. The results show that our proposed multitask model outperforms the baseline Open-Unmix model on the mixture of Mixing Secrets and MedleyDB dataset while maintaining comparable performance on the MUSDB dataset.
Recent approaches for music source separation are almost exclusively based on deep neural networks, mostly employing recurrent neural networks (RNNs). Although RNNs are in many cases superior than other types of deep neural networks for sequence processing, they are known to have specific difficulties in training and parallelization, especially for the typically long sequences encountered in music source separation. In this paper we present a use-case of replacing RNNs with depth-wise separable (DWS) convolutions, which are a lightweight and faster variant of the typical convolutions. We focus on singing voice separation, employing an RNN architecture, and we replace the RNNs with DWS convolutions (DWS-CNNs). We conduct an ablation study and examine the effect of the number of channels and layers of DWS-CNNs on the source separation performance, by utilizing the standard metrics of signal-to-artifacts, signal-to-interference, and signal-to-distortion ratio. Our results show that by replacing RNNs with DWS-CNNs yields an improvement of 1.20, 0.06, 0.37 dB, respectively, while using only 20.57% of the amount of parameters of the RNN architecture.
A fairly straightforward approach for music source separation is to train independent models, wherein each model is dedicated for estimating only a specific source. Training a single model to estimate multiple sources generally does not perform as well as the independent dedicated models. However, Conditioned U-Net (C-U-Net) uses a control mechanism to train a single model for multi-source separation and attempts to achieve a performance comparable to that of the dedicated models. We propose a multi-channel U-Net (M-U-Net) trained using a weighted multi-task loss as an alternative to the C-U-Net. We investigate two weighting strategies for our multi-task loss: 1) Dynamic Weighted Average (DWA), and 2) Energy Based Weighting (EBW). DWA determines the weights by tracking the rate of change of loss of each task during training. EBW aims to neutralize the effect of the training bias arising from the difference in energy levels of each of the sources in a mixture. Our methods provide three-fold advantages compared to C-UNet: 1) Fewer effective training iterations per epoch, 2) Fewer trainable network parameters (no control parameters), and 3) Faster processing at inference. Our methods achieve performance comparable to that of C-U-Net and the dedicated U-Nets at a much lower training cost.
Music source separation has been a popular topic in signal processing for decades, not only because of its technical difficulty, but also due to its importance to many commercial applications, such as automatic karoake and remixing. In this work, we propose a novel self-attention network to separate voice and accompaniment in music. First, a convolutional neural network (CNN) with densely-connected CNN blocks is built as our base network. We then insert self-attention subnets at different levels of the base CNN to make use of the long-term intra-dependency of music, i.e., repetition. Within self-attention subnets, repetitions of the same musical patterns inform reconstruction of other repetitions, for better source separation performance. Results show the proposed method leads to 19.5% relative improvement in vocals separation in terms of SDR. We compare our methods with state-of-the-art systems i.e. MMDenseNet and MMDenseLSTM.