We introduce two unsupervised source separation methods, which involve self-supervised training from single-channel two-source speech mixtures. Our first method, mixture permutation invariant training (MixPIT), enables learning a neural network model which separates the underlying sources via a challenging proxy task without supervision from the reference sources. Our second method, cyclic mixture permutation invariant training (MixCycle), uses MixPIT as a building block in a cyclic fashion for continuous learning. MixCycle gradually converts the problem from separating mixtures of mixtures into separating single mixtures. We compare our methods to common supervised and unsupervised baselines: permutation invariant training with dynamic mixing (PIT-DM) and mixture invariant training (MixIT). We show that MixCycle outperforms MixIT and reaches a performance level very close to the supervised baseline (PIT-DM) while circumventing the over-separation issue of MixIT. Also, we propose a self-evaluation technique inspired by MixCycle that estimates model performance without utilizing any reference sources. We show that it yields results consistent with an evaluation on reference sources (LibriMix) and also with an informal listening test conducted on a real-life mixtures dataset (REAL-M).

In this letter, we propose a source separation method that is trained by observing the mixtures and the class labels of the sources present in the mixture without any access to isolated sources. Since our method does not require source class labels for every time-frequency bin but only a single label for each source constituting the mixture signal, we call this scenario as weak class supervision. We associate a variational autoencoder (VAE) with each source class within a nonnegative (compositional) model. Each VAE provides a prior model to identify the signal from its associated class in a sound mixture. After training the model on mixtures, we obtain a generative model for each source class and demonstrate our method on one-second mixtures of utterances of digits from 0 to 9. We show that the separation performance obtained by source class supervision is as good as the performance obtained by source signal supervision.

In this paper, we propose an adaptive time-frequency resolution based single channel sound source separation method using Non-negative Tensor Factorization (NTF). The model aims to alleviate drawbacks of working by fixed length Short Time Fourier Transform (STFT) by minimizing the smearing of signal energy in both time and frequency. A joint optimization scheme has been applied based on KL-divergence where each layer of the tensor represents the mixture at a different resolution. In order to enclose sparseness into factorization, the resynthesis is made through an adaptive weighted fusion procedure which combines the separated sources in a manner that maximizes the energy concentration. Test results reported over a large sound database indicate the introduced NTF based fusion method improves the sound quality both in terms of conventional and perceptual distortion measures.

In this paper, we propose an adaptive time-frequency resolution approach for the single channel source separation problem. The aim is to improve the quality and intelligibility of the separated sources by adapting the time-frequency resolution of the analysis window to the characteristic of the signal under consideration. The results evaluated on a large test set show the improvements obtained by the proposed algorithm.

In this paper we develop a probabilistic interpretation and a full Bayesian inference for non-negative matrix deconvolution (NMFD) model. Our ultimate goal is unsupervised extraction of multiple sound objects from a single channel auditory scene. The proposed method facilitates automatic model selection and determination of the sparsity criteria. Our approach retains attractive features of standard NMFD based methods such as fast convergence and easy implementation. We demonstrate the use of this algorithm in the log-frequency magnitude spectrum domain, where we employ it to perform model order selection and control sparseness directly.

In contrast to the fixed rate modeling of the conventional methods, recently introduced variable rate particle filters (VRPF) achieves to track maneuvering objects with a small number of states by imposing a probability distribution on state arrival times. Although this enables VRPF an appealing method, representing the target motion dynamics with a single model hinders the capability of estimating maneuver parameters precisely. To overcome this weakness we have incorporated multiple model approach with the variable rate model structure. The introduced model referred as Multiple Model Variable Rate Particle Filter (MM-VRPF) utilizes a parsimonious representation for smooth regions of trajectory while it adaptively locates frequent state points at high maneuver regions, resulting in a much more accurate tracking. Simulation results obtained in a bearings-only target tracking problem show that the proposed model outperforms the conventional VRPF, the fixed rate multiple model particle filters (MMPF) and interacting multiple model using extended Kalman filters (IMM-EKF).

We propose a method for embedding a robust forensic tracking watermark within the AAC decoder during audio playback. For predefined frequency bands, the method intercepts and modifies the scale-factors, which are present in the AAC bit-stream. It thereby modulates the short-time envelope of the band-limited audio and embeds a watermark which is robust to various attacks such as recompression and acoustic transmission. Due to its low complexity - a few additions per AAC frame - the method is suitable for implementation even in resource constrained devices such as portable players. Furthermore, decode-time embedding enables dynamic payload assignments which may be used to signal playback time and location.

In digital rights-management systems, forensic watermarking complements encryption and deters the capture and unauthorized redistribution of the rendered content. In this paper, we propose a novel watermarking method which is integrated into the advanced audio coding (AAC) standard's decoding process. For predefined frequency bands, the method intercepts and modifies the scale factors, which are utilized for dequantization of spectral coefficients. It thereby modulates the short-time envelope of the bandlimited audio and embeds a watermark which is robust to various attacks, such as capture with a microphone and recompression at lower bit rates. Inclusion of watermark embedding in the AAC decoder has practically no effect on the decoding complexity. As a result, the proposed method can be integrated even into resource-constrained devices, such as portable players without any additional hardware.

Most of the watermark (WM) decoding schemes use correlation-based methods because of their simplicity. In these methods, the WM signal embedded through a secret key is assumed as uncorrelated with the host signal. This is a hard restriction that can never be achieved and correlation between the received signal and the secret key becomes greater than zero even though the received signal is un-watermarked. Mostly a decision threshold specified semi-automatically is used at the decoding site. Since the audio watermarking is a nonlinear process that guarantees the inaudibility, there is no analytic way of determining an optimal threshold value that makes the WM decoding problem harder. This paper introduces a learning scheme followed by a nonlinear classification thus eliminates the threshold specification problem. The decoding process is modelled as a three-class classification problem and support vector machines (SVMs) are used in the learning of the embedded data. The decoding and detection performances of the developed system are greater than 98% and 95%, respectively. When the watermark-to-signal-ratio (WSR) is higher than -30 dB, system false alarm ratios remain less than 2%. It is shown that the introduced WM decoding method is robust to additive noise and most of add/remove and filter attacks of Stirmark