arXiv subjects

Pavel Andreev

Publications and source records attributed to Pavel Andreev.

13 recordsLinked to original sources

arXiv2024

FINALLY: fast and universal speech enhancement with studio-like quality

In this paper, we address the challenge of speech enhancement in real-world recordings, which often contain various forms of distortion, such as background noise, reverberation, and microphone artifacts. We revisit the use of Generative Adversarial Networks (GANs) for speech enhancement and theoretically show that GANs are naturally inclined to seek the point of maximum density within the conditional clean speech distribution, which, as we argue, is essential for the speech enhancement task. We study various feature extractors for perceptual loss to facilitate the stability of adversarial training, developing a methodology for probing the structure of the feature space. This leads us to integrate WavLM-based perceptual loss into MS-STFT adversarial training pipeline, creating an effective and stable training procedure for the speech enhancement model. The resulting speech enhancement model, which we refer to as FINALLY, builds upon the HiFi++ architecture, augmented with a WavLM encoder and a novel training pipeline. Empirical results on various datasets confirm our model's ability to produce clear, high-quality speech at 48 kHz, achieving state-of-the-art performance in the field of speech enhancement. Demo page: https://samsunglabs.github.io/FINALLY-page

cs.SD↗

arXiv2024

Search for manifestations of spin-torsion coupling

We investigate the axial vector torsion-spin coupling effects in the framework of the Poincaré gauge theory of gravity with the general Yang-Mills type Lagrangian. The dynamical equations for the ``electric'' and ``magnetic'' components of the torsion field variable are obtained in the general form and it is shown that the helicity density and the spin density of the electromagnetic field appear as the physical sources. The modified Maxwell's equations for the electromagnetic field are derived, and the electromagnetic wave propagation under the action of the uniform homogeneous torsion field is considered. We demonstrate the Faraday effect of rotation of the polarization for such a wave, and establish the strong bound on the possible cosmic axial torsion field from the astrophysical data.

gr-qc↗

arXiv2024

Speech Boosting: Low-Latency Live Speech Enhancement for TWS Earbuds

This paper introduces a speech enhancement solution tailored for true wireless stereo (TWS) earbuds on-device usage. The solution was specifically designed to support conversations in noisy environments, with active noise cancellation (ANC) activated. The primary challenges for speech enhancement models in this context arise from computational complexity that limits on-device usage and latency that must be less than 3 ms to preserve a live conversation. To address these issues, we evaluated several crucial design elements, including the network architecture and domain, design of loss functions, pruning method, and hardware-specific optimization. Consequently, we demonstrated substantial improvements in speech enhancement quality compared with that in baseline models, while simultaneously reducing the computational complexity and algorithmic latency.

eess.AS↗

arXiv2024

Investigation of Microstructure and Corrosion Resistance of Ti-Al-V Titanium Alloys Obtained by Spark Plasma Sintering

The research results of the microstructure and corrosion resistance of Ti and Ti-Al-V Russian industrial titanium alloys obtained by spark plasma sintering (SPS) are described. Investigations of the microstructure, phase composition, hardness, tensile strength, electrochemical corrosion resistance and hot salt corrosion of Ti-Al-V titanium alloy specimens were carried out. It was shown that the alloy specimens have a uniform highly dense microstructure and high hardness values. The studied alloys also have high resistance to electrochemical corrosion during tests in acidic aqueous solution causing the intergranular corrosion as well as high resistance to the hot salt corrosion. The assumption that the high hardness of the alloys as well as the differences in the corrosion resistance of the central and lateral parts of the specimens are due to the diffusion of carbon from the graphite mold into the specimen surface was suggested.

physics.app-ph↗

arXiv2024

A new microscopic representation of the spin dynamics in quantum systems with the Coulomb exchange interactions

There is a version of the Landau-Lifshitz equation that takes into account the Coulomb exchange interactions between atoms, expressed by the term $\sim\bm{s}\times\triangle\bm{s}$. On the other hand, ions in the magnetic materials have several valence electrons on the $d$-shell, and therefore the Hamiltonian of many-electron atoms with spins $S>1$ should include a biquadratic exchange interaction. We first propose a new fundamental microscopic derivation of the spin density evolution equation with an explicit form of biquadratic exchange interaction using the method of many-particle quantum hydrodynamics. The equation for the evolution of the spin density is obtained from the many-particle Schrodinger-Pauli equation and contains the contributions of the usual Coulomb exchange interaction and the biquadratic exchange. Furthermore, the derived biquadratic exchange torque in the spin density evolution equation is proportional to the nematic tensor for the medium of atoms with spin $\textit{S = 1}$. Our method may be very attractive for further studies of the magnetoelectric effect in multiferroics.

cond-mat.mtrl-sci↗

arXiv2023

HiFi++: a Unified Framework for Bandwidth Extension and Speech Enhancement

Generative adversarial networks have recently demonstrated outstanding performance in neural vocoding outperforming best autoregressive and flow-based models. In this paper, we show that this success can be extended to other tasks of conditional audio generation. In particular, building upon HiFi vocoders, we propose a novel HiFi++ general framework for bandwidth extension and speech enhancement. We show that with the improved generator architecture, HiFi++ performs better or comparably with the state-of-the-art in these tasks while spending significantly less computational resources. The effectiveness of our approach is validated through a series of extensive experiments.

cs.SD↗

arXiv2023

Iterative autoregression: a novel trick to improve your low-latency speech enhancement model

Streaming models are an essential component of real-time speech enhancement tools. The streaming regime constrains speech enhancement models to use only a tiny context of future information. As a result, the low-latency streaming setup is generally considered a challenging task and has a significant negative impact on the model's quality. However, the sequential nature of streaming generation offers a natural possibility for autoregression, that is, utilizing previous predictions while making current ones. The conventional method for training autoregressive models is teacher forcing, but its primary drawback lies in the training-inference mismatch that can lead to a substantial degradation in quality. In this study, we propose a straightforward yet effective alternative technique for training autoregressive low-latency speech enhancement models. We demonstrate that the proposed approach leads to stable improvement across diverse architectures and training scenarios.

cs.SD↗

arXiv2023

UnDiff: Unsupervised Voice Restoration with Unconditional Diffusion Model

This paper introduces UnDiff, a diffusion probabilistic model capable of solving various speech inverse tasks. Being once trained for speech waveform generation in an unconditional manner, it can be adapted to different tasks including degradation inversion, neural vocoding, and source separation. In this paper, we, first, tackle the challenging problem of unconditional waveform generation by comparing different neural architectures and preconditioning domains. After that, we demonstrate how the trained unconditional diffusion could be adapted to different tasks of speech processing by the means of recent developments in post-training conditioning of diffusion models. Finally, we demonstrate the performance of the proposed technique on the tasks of bandwidth extension, declipping, vocoding, and speech source separation and compare it to the baselines. The codes are publicly available.

cs.SD↗

arXiv2022

FFC-SE: Fast Fourier Convolution for Speech Enhancement

Fast Fourier convolution (FFC) is the recently proposed neural operator showing promising performance in several computer vision problems. The FFC operator allows employing large receptive field operations within early layers of the neural network. It was shown to be especially helpful for inpainting of periodic structures which are common in audio processing. In this work, we design neural network architectures which adapt FFC for speech enhancement. We hypothesize that a large receptive field allows these networks to produce more coherent phases than vanilla convolutional models, and validate this hypothesis experimentally. We found that neural networks based on Fast Fourier convolution outperform analogous convolutional models and show better or comparable results with other speech enhancement baselines.

cs.SD↗

arXiv2021

Quantization of Generative Adversarial Networks for Efficient Inference: a Methodological Study

Generative adversarial networks (GANs) have an enormous potential impact on digital content creation, e.g., photo-realistic digital avatars, semantic content editing, and quality enhancement of speech and images. However, the performance of modern GANs comes together with massive amounts of computations performed during the inference and high energy consumption. That complicates, or even makes impossible, their deployment on edge devices. The problem can be reduced with quantization -- a neural network compression technique that facilitates hardware-friendly inference by replacing floating-point computations with low-bit integer ones. While quantization is well established for discriminative models, the performance of modern quantization techniques in application to GANs remains unclear. GANs generate content of a more complex structure than discriminative models, and thus quantization of GANs is significantly more challenging. To tackle this problem, we perform an extensive experimental study of state-of-art quantization techniques on three diverse GAN architectures, namely StyleGAN, Self-Attention GAN, and CycleGAN. As a result, we discovered practical recipes that allowed us to successfully quantize these models for inference with 4/8-bit weights and 8-bit activations while preserving the quality of the original full-precision models.

cs.AI↗

arXiv2021

Hydrodynamic description of Weyl fermions in condensed state of matter

Due to the many unique transport properties of Weyl semimetals, they are promising materials for modern electronics. We investigate the electrons in the strong coupling approximation near Weyl points based on their representation as massless Weyl fermions. We have constructed a new fluid model based on the many-particle quantum hydrodynamics method to describe the behavior of electrons gas with different chirality near Weyl points in the low-energy limit in the external electromagnetic fields, based on the many-particle Weyl equation and many-particle wave function. The derived system of equations forms a closed apparatus for describing the dynamics of the electron current, spin density and spin current density. Based on the proposed model, we considered small perturbations in the Weyl fermion system in an external uniform magnetic field and predicted the new type of eigenwaves in the systems of the electrons near the Weyl points.

cond-mat.mes-hall↗

arXiv2014

Semi-relativistic hydrodynamics of three-dimensional and low-dimensional quantum plasma

Contributions of the current-current and Darwin interactions and weak-relativistic addition to kinetic energy in the quantum hydrodynamic equations are considered. Features of hydrodynamic equations for two-dimensional layer of plasma (two-dimensional electron gas for instance) are described. It is shown that the force fields caused by the Darwin interaction and weak-relativistic addition to kinetic energy are partially reduced. Dispersion of three- and two-dimensional semi-relativistic Langmuir waves is calculated.

physics.plasm-ph↗

arXiv2004

A characterization of isometries of CAT(0)-space as maps preserving diagonal tube

We give positive answers for questions by Berestovskii. Namely, we prove that every bijection of locally compact geodesically complete and connected at infinity CAT(0)-space $X$ onto itself preserving some fixed distance or satellite relations is an isometry of this space. The proof of this theorem is based on another result stated by Berestovskii as a problem: the metric of the space $X$ may be recovered from its diagonal tube corresponding to an arbitrary number $r > 0$.

math.GT↗