arXiv subjects

Stanislav Yu. Kruchinin

Publications and source records attributed to Stanislav Yu. Kruchinin.

13 recordsLinked to original sources

arXiv2026

Decaf: A privacy preserving speech codec using speaker disentanglement and canonical voice conversion

We present DECAF, a privacy preserving neural speech codec that obfuscates a speaker's voice while preserving linguistic content while maintaining automatic speech recognition (ASR) performance at very low bitrates, inspired by decaffeination. At the transmitter end, speech is encoded into speaker independent content embeddings, which are compressed using residual vector quantization and transmitted without any speaker related information. At the receiver, a canonical speaker embedding, shared a priori between endpoints, is used for waveform reconstruction, enabling deterministic and consistent obfuscation of a speaker's voice. The proposed framework leverages an information bottleneck applied to self supervised representations, along with a separate speaker embedding branch, to achieve effective speaker content disentanglement. We further incorporate a CTC-based auxiliary objective, encouraging content representations that are well aligned with downstream ASR tasks. We show that DECAF operating at a bit rate of 0.5 kbps achieves an Equal Error Rate (EER) of up to 43.5% for a speaker verification system, while maintaining competitive ASR performance, yielding a relative reduction in word error rate of 33.2% compared to a state of the art method.

eess.AS↗

arXiv2026

Self-Consistent Spectral Quadrature Approach to Many-Body Green Functions

We develop a self-consistent spectral quadrature (sc-SQ) framework for the calculation of many-body Green functions. The method approximates the Källén--Lehmann spectral measure by Gauss--Christoffel (GC) quadrature, yielding a rational Green function representation with guaranteed spectral positivity that exactly reproduces the first $2N$ spectral moments at pole order $N$. A key component is an SVD-based rank-selection criterion on the Hankel matrix, which identifies the numerically resolvable pole rank $N^*$ from the singular-value gap and acts as a precision-guided diagnostic of correlation complexity. The scheme is made self-consistent by requiring that the spectral function used to evaluate expectation values coincides with the spectral function generated by the quadrature reconstruction. This defines a fixed-point hierarchy that connects systematically to established approximations, including Hartree--Fock and Hubbard-I, and incorporates non-perturbative features such as multi-peak spectral structure. We benchmark the approach for the Anderson impurity model against numerical renormalization group (NRG) results and apply it within dynamical mean-field theory for the Hubbard model on the Bethe lattice. The method captures the three-peak Anderson impurity spectrum and the suppression of quasiparticle weight in the half-filled Hubbard model on the Bethe lattice, including Mott-gap formation on the insulating branch for $N\geqslant 5$, in qualitative agreement with NRG references.

cond-mat.str-el↗

arXiv2024

XANE Background Acoustic Embeddings: Ablation and Clustering Analysis

We explore the recently proposed explainable acoustic neural embedding~(XANE) system that models the background acoustics of a speech signal in a non-intrusive manner. The XANE embeddings are used to estimate specific parameters related to the background acoustic properties of the signal which allows the embeddings to be explainable in terms of those parameters. We perform ablation studies on the XANE system and show that estimating all acoustic parameters jointly has an overall positive effect. Furthermore, we illustrate the value of XANE embeddings by performing clustering experiments on unseen test data and show that the proposed embeddings achieve a mean F1 score of 92\% for three different tasks, outperforming significantly the WavLM based signal embeddings and are complimentary to speaker embeddings.

eess.AS↗

arXiv2023

Excitons in mesoscopically reconstructed moiré heterostructures

Moiré effects in twisted or lattice-incommensurate vertical assemblies of two-dimensional crystals give rise to a new class of quantum materials with rich transport and optical phenomena, including correlated electron physics in flat bands of bilayer graphene and moiré excitons in semiconductor heterostructures. These phenomena arise from modulations of interlayer hybridization on the nanoscale of spatially varying atomic registries of moiré supercells. Due to finite elasticity, however, lattices of marginally-twisted homobilayers and heterostructures can transform from moiré to periodically reconstructed patterns with triangular or hexagonal tiling. Here, we expand the notion of nanoscale lattice reconstruction to the mesoscopic scale of extended samples and demonstrate rich consequences in optical studies of excitons in MoSe$_2$-WSe$_2$ heterostructures with parallel and antiparallel alignment. Our results provide a unified perspective on diverse and partly controversial signatures of moiré excitons in semiconductor heterostructures by identifying domains with exciton properties of distinct effective dimensionality and establish mesoscopic reconstruction as a compelling feature of real samples and devices with inherent finite-size effects and disorder. Generalized to stacks of other two-dimensional materials, this notion of mesoscale domain formation with emergent topological defects and percolation networks will instructively expand our understanding of fundamental electronic, optical, and magnetic properties of van der Waals heterostructures.

cond-mat.mes-hall↗

arXiv2022

Spatial Processing Front-End For Distant ASR Exploiting Self-Attention Channel Combinator

We present a novel multi-channel front-end based on channel shortening with theWeighted Prediction Error (WPE) method followed by a fixed MVDR beamformer used in combination with a recently proposed self-attention-based channel combination (SACC) scheme, for tackling the distant ASR problem. We show that the proposed system used as part of a ContextNet based end-to-end (E2E) ASR system outperforms leading ASR systems as demonstrated by a 21.6% reduction in relative WER on a multi-channel LibriSpeech playback dataset. We also show how dereverberation prior to beamforming is beneficial and compare the WPE method with a modified neural channel shortening approach. An analysis of the non-intrusive estimate of the signal C50 confirms that the 8 channel WPE method provides significant dereverberation of the signals (13.6 dB improvement). We also show how the weights of the SACC system allow the extraction of accurate spatial information which can be beneficial for other speech processing applications like diarization.

eess.AS↗

arXiv2021

Moiré excitons in MoSe$_2$-WSe$_2$ heterobilayers and heterotrilayers

Layered two-dimensional materials exhibit rich transport and optical phenomena in twisted or lattice-incommensurate heterostructures with spatial variations of interlayer hybridization arising from moiré interference effects. Here, we report experimental and theoretical studies of excitons in twisted heterobilayers and heterotrilayers of transition metal dichalcogenides. Using MoSe$_2$-WSe$_2$ stacks as representative realizations of twisted van der Waals bilayer and trilayer heterostructures, we observe contrasting optical signatures and interpret them in the theoretical framework of interlayer moiré excitons in different spin and valley configurations. We conclude that the photoluminescence of MoSe$_2$-WSe$_2$ heterobilayer is consistent with joint contributions from radiatively decaying valley-direct interlayer excitons and phonon-assisted emission from momentum-indirect reservoirs that reside in spatially distinct regions of moiré supercells, whereas the heterotrilayer emission is entirely due to momentum-dark interlayer excitons of hybrid-layer valleys. Our results highlight the profound role of interlayer hybridization for transition metal dichalcogenide heterostacks and other realizations of multi-layered semiconductor van der Waals heterostructures.

cond-mat.mes-hall↗

arXiv2019

Giant Stokes shifts in AgInS$_2$ nanocrystals with trapped charge carriers

Nanocrystals of AgInS$_2$ demonstrate giant Stokes shifts ~ 1 eV, the nature of which is still not clearly understood. We propose a theoretical model of this phenomenon bringing together several different mechanisms previously considered only separately. We take into account the contribution of electron-electron interaction with the hybrid density functional theory, as well as the renormalization of energy spectrum due to the electron-phonon coupling. Furthermore, we consider the presence of at least one point defect responsible for hole trapping and the formation of a localized polaron state. Our numerical simulations show that photoluminescence due to the recombination of a non-trapped electron and a trapped hole results in the giant Stokes shift in AgInS$_2$ nanocrystal, which is in close agreement with the recent experimental results.

cond-mat.mes-hall↗

arXiv2018

Strong-field Phenomena in Periodic Systems

The advent of visible-infrared laser pulses carrying a substantial fraction of their energy in a single field oscillation cycle has opened a new era in the experimental investigation of ultrafast processes in semiconductors and dielectrics (bulk as well as nanostructured), motivated by the quest for the ultimate frontiers of electron-based signal metrology and processing. Exploring ways to approach those frontiers requires insight into the physics underlying the interaction of strong high-frequency (optical) fields with electrons moving in periodic potentials. This Colloquium aims at providing this insight. Introduction to the foundations of strong-field phenomena defines and compares regimes of field--matter interaction in periodic systems, including (perfect) crystals as well as optical and semiconductor superlattices, followed by a review of recent experimental advances in the study of strong-field dynamics in crystals and nanostructures. Avenues toward measuring and controlling electronic processes up to petahertz frequencies are discussed.

quant-ph↗

arXiv2016

Sub-cycle optical control of current in a semiconductor: from the multiphoton to the tunneling regime

Nonlinear interactions between ultrashort optical waveforms and solids can be used to induce and steer electric current on a femtosecond (fs) timescale, holding promise for electronic signal processing at PHz frequencies [Nature 493, 70 (2013)]. So far, this approach has been limited to insulators, requiring extremely strong peak electric fields and intensities. Here, we show all-optical generation and control of directly measurable electric current in a semiconductor relevant for high-speed and high-power (opto)electronics, gallium nitride (GaN), within an optical cycle and on a timescale shorter than 2 fs, at intensities at least an order of magnitude lower than those required for dielectrics. Our approach opens the door to PHz electronics and metrology, applicable to low-power (non-amplified) laser pulses, and may lead to future applications in semiconductor and photonic integrated circuit technologies.

cond-mat.mes-hall↗

arXiv2015

Strong-Field Resonant Dynamics in Semiconductors

We predict that a direct bandgap semiconductor (GaAs) resonantly excited by a strong ultrashort laser pulse exhibits a novel regime: kicked anharmonic Rabi oscillations (KARO). In this regime, Rabi oscillations are strongly coupled to intraband motion, and interband transitions mainly take place during short times when electrons pass near the Brillouin zone center where electron populations undergo very rapid changes. Asymmetry of the residual population distribution induces an electric current controlled by the carrier-envelope phase. The predicted effects are experimentally observable using photoemission and terahertz spectroscopies.

cond-mat.mes-hall↗

arXiv2015

Ultrafast control of strong-field electron dynamics in solids

We review theoretical foundations and some recent progress related to the quest of controlling the motion of charge carriers with intense laser pulses and optical waveforms. The tools and techniques of attosecond science enable detailed investigations of a relatively unexplored regime of nondestructive strong-field effects. Such extremely nonlinear effects may be utilized to steer electron motion with precisely controlled optical fields and switch electric currents at a rate that is far beyond the capabilities of conventional electronics.

cond-mat.mes-hall↗

arXiv2013

Theory of strong-field injection and control of photocurrent in dielectrics and wide bandgap semiconductors

We propose a theory of optically-induced currents in dielectrics and wide-gap semiconductors exposed to a non-resonant ultrashort laser pulse with a stabilized carrier-envelope phase. In order to describe strong-field electron dynamics, equations for density matrix have been solved self-consistently with equations for the macroscopic electric field inside the medium, which we model by a one-dimensional potential. We provide a detailed analysis of physically important quantities (band populations, macroscopic polarization, and transferred charge), which reveals that carrier-envelope phase control of the electric current can be interpreted as a result of quantum-mechanical interference of multiphoton excitation channels. Our numerical results are in good agreement with experimental data.

cond-mat.mtrl-sci↗

arXiv2013

Quantum beats in the polarization response of a dielectric to intense few-cycle laser pulses

We have investigated the polarization response of a dielectric to intense few-cycle laser pulses with a focus on interband tunnelling. Once charge carriers are created in an initially empty conduction band, they make a significant contribution to the polarization response. In particular, the coherent superposition of conduction- and valence-band states results in quantum beats. This quantum-beat part of the polarization response is affected by the excitation dynamics and attosecond-scale motion of charge carriers in an intense laser field. Our analysis shows that, with the onset of Bloch oscillations or tunnelling, the nonlinear polarization response becomes sensitive to the carrier-envelope phase of the laser pulse.

physics.atom-ph↗