On phase recovery and preserving early reflections for deep-learning speech dereverberation.

In indoor environments, reverberation often distorts clean speech. Although deep learning-based speech dereverberation approaches have shown much better performance than traditional ones, the inferior speech quality of the dereverberated speech caused by magnitude distortion and limited phase recovery is still a serious problem for practical applications. This paper improves the performance of deep learning-based speech dereverberation from the perspectives of both network design and mapping target optimization. Specifically, on the one hand, a bifurcated-and-fusion network and its guidance loss functions were designed to help reduce the magnitude distortion while enhancing the phase recovery. On the other hand, the time boundary between the early and late reflections in the mapped speech was investigated, so as to make a balance between the reverberation tailing effect and the difficulty of magnitude/phase recovery. Mathematical derivations were provided to show the rationality of the specially designed loss functions. Geometric illustrations were given to explain the importance of preserving early reflections in reducing the difficulty of phase recovery. Ablation study results confirmed the validity of the proposed network topology and the importance of preserving 20 ms early reflections in the mapped speech. Objective and subjective test results showed that the proposed system outperformed other baselines in the speech dereverberation task.

Giant enhancement of second-harmonic generation of indium selenide on planar Au.

Two-dimensional (2D) van der Waals layered γ-type indium selenide (γ-InSe) holds great promise for the development of ultrathin and low-energy-consumption nonlinear optical devices due to its broken inversion symmetry regardless of layer number. Nevertheless, the 2D InSe thin flakes still exhibit short light-matter interaction lengths, thus resulting in low efficiencies of nonlinear optical processes. In this work, we provide a facile 2D semiconductor-metal structure consisting of InSe thin flakes (thickness: 11-54 nm) on planar Au film, which exhibits great second-harmonic generation (SHG) enhancement by a factor of up to 1182. The SHG enhancement is attributed to the interference effect-induced strong electric field in highly absorbing InSe; meanwhile, the increase in reflectivity by Au film also plays an important role. Furthermore, the InSe thickness and excitation wavelength dependences of enhancement factors are revealed. This work provides a convenient approach to developing high-efficiency 2D nonlinear optical devices with ultrathin form.

Giant Out-of-Plane Exciton Emission Enhancement in Two-Dimensional Indium Selenide via a Plasmonic Nanocavity.

Out-of-plane (OP) exciton-based emitters in two-dimensional semiconductor materials are attractive candidates for novel photonic applications, such as radially polarized sources, integrated photonic chips, and quantum communications. However, their low quantum efficiency resulting from forbidden transitions limits their practicality. In this work, we achieve a giant enhancement of up to 34000 for OP exciton emission in indium selenide (InSe) via a designed Ag nanocube-over-Au film plasmonic nanocavity. The large photoluminescence enhancement factor (PLEF) is attributed to the induced OP local electric field (Ez) within the nanocavity, which facilitates effective OP exciton-plasmon interaction and subsequent tremendous enhancement. Moreover, the nanoantenna effect resulting from the effective interaction improves the directivity of spontaneous radiation. Our results not only reveal an effective photoluminescence enhancement approach for OP excitons but also present an avenue for designing on-chip photonic devices with an OP dipole orientation.

Low-latency monaural speech enhancement with deep filter-bank equalizer.

It is highly desirable that speech enhancement algorithms can achieve good performance while keeping low latency for many applications, such as digital hearing aids, mobile phones, acoustically transparent hearing devices, and public address systems. To improve the performance of traditional low-latency speech enhancement algorithms, a deep filter-bank equalizer (FBE) framework was proposed that integrated a deep learning-based subband noise reduction network with a deep learning-based shortened digital filter mapping network. In the first network, a deep learning model was trained with a controllable small frame shift to satisfy the low-latency demand, i.e., no greater than 4 ms, so as to obtain (complex) subband gains that could be regarded as an adaptive digital filter in each frame. In the second network, to reduce the latency, this adaptive digital filter was implicitly shortened by a deep learning-based framework and was then applied to noisy speech to reconstruct the enhanced speech without the overlap-add method. Experimental results on the WSJ0-SI84 corpus indicated that the proposed DeepFBE with only 4-ms latency achieved much better performance than traditional low-latency speech enhancement algorithms across several objective metrics. Listening test results further confirmed that our approach achieved higher speech quality than other methods.

Visible to near-infrared photodetector with novel optoelectronic performance based on graphene/S-doped InSe heterostructure on h-BN substrate.

van der Waals heterostructures of two-dimensional (2D) materials have attracted considerable attention due to their flexibility in the design of new functional devices. Despite numerous studies on graphene/2D semiconductor heterostructures, their optoelectronic applications are significantly hindered because of several disadvantages, such as large band gaps and chemical instability. In this work, we demonstrate the fabrication of graphene/S-doped InSe heterostructure photodetectors with excellent photoresponse performance, and this is attributed to the moderate band gap and band gap engineering by element doping of InSe as well as the high carrier mobility of graphene. In particular, the graphene/InSe0.9S0.1 device achieves an ultrahigh photoresponsivity of ∼4.9 × 106 A W-1 at 700 nm and an EQE of 8.7 × 108%, and it exhibits broadband photodetection (visible to near-infrared). More importantly, by virtue of the interaction between n-type graphene arising from the influence of h-BN as a dielectric layer and S-doped InSe with a high work-function, our devices always exhibited positive photocurrent when the polarity of the gate voltage is adjusted, and is different from that the previously reported graphene/2D semiconductor photodetectors. This work not only provides a promising platform for highly efficient broadband photodetectors but also sheds light on tuning the optoelectronic performance through band gap engineering and designing novel heterostructures-based various 2D materials.

Surface-Modified Ultrathin InSe Nanosheets with Enhanced Stability and Photoluminescence for High-Performance Optoelectronics.

Indium selenide (InSe) has become a research hotspot because of its favorable carrier mobility and thickness-tunable band gap, showing great application potential in high-performance optoelectronic devices. The trend of miniaturization in optoelectronics has forced the feature sizes of the electronic components to shrink accordingly. Therefore, atomically thin InSe crystals may play an important role in future optoelectronics. Given the instability and ultralow photoluminescent (PL) emission of mechanically exfoliated ultrathin InSe, synthesis of highly stable mono- and few-layer InSe nanosheets with high PL efficiency has become crucial. Herein, ultrathin InSe nanosheets were prepared via thermal annealing of electrochemically intercalated products from bulk InSe. The size and yield of the as-prepared nanosheets were up to ∼160 µm and ∼70%, respectively, and ∼80% of the nanosheets were less than five layer. Impressively, the as-prepared nanosheets showed greatly enhanced stability and PL emission because of surface modification by carbon species. Efficient photoresponsivity of 2 A/W was achieved in the as-prepared nanosheet-based devices. These nanosheets were further assembled into large-area thin films with photoresponsivity of 16 A/W and an average Hall mobility of about 5 cm2 V-1 s-1. Finally, one-dimensional (1D) InSe nanoscrolls with a length up to 90 µm were constructed by solvent-assisted self-assembly of the exfoliated nanosheets.

Synthesis of large-area uniform Si2Te3 thin films for p-type electronic devices.

Two-dimensional (2D) p-n junctions are basic components of various functional devices. However, the shortage of natural p-type 2D semiconductors makes it difficult to achieve both p-type and n-type transport in high-performance multifunctional devices. Here, continuous and uniform p-type Si2Te3 thin films are grown on SiO2/Si substrates, which are simultaneously used as an in situ Si source. Large-size 2D films with dimensions of ∼8 × 2 cm2 are prepared for the first time using a reliable and simple chemical vapor deposition (CVD) technique. Film growth occurs via the vapor-liquid-solid mechanism, allowing the film thickness to be controlled by the substrate temperature. As the Si2Te3 film thickness increases from 3 to 8 nm, the bandgap decreases from 2.07 to 1.65 eV. Moreover, the directly grown thin films possess high crystallinity, showing electronic properties that are comparable to those of MoTe2 crystals and MoS2 films. Therefore, this large-area growth of p-type Si2Te3 enriches the 2D semiconductor library and opens up a new platform for the study of p-type Si2Te3, which has potential for application in p-n junctions.

Joint estimation of binaural distance and azimuth by exploiting deep neural networks.

The state-of-the-art supervised binaural distance estimation methods often use binaural features that are related to both the distance and the azimuth, and thus the distance estimation accuracy may degrade a great deal with fluctuant azimuth. To incorporate the azimuth on estimating the distance, this paper proposes a supervised method to jointly estimate the azimuth and the distance of binaural signals based on deep neural networks (DNNs). In this method, the subband binaural features, including many statistical properties of several subband binaural features and the binaural spectral magnitude difference standard deviation, are extracted together as cues to jointly estimate the azimuth and the distance using binaural signals by exploiting a multi-objective DNN framework. Especially, both the azimuth and the distance cues are utilized in the learning stage of the error back-propagation in the multi-objective DNN framework, which can improve the generalization ability of the azimuth and the distance estimation. Experimental results demonstrate that the proposed method can not only achieve high azimuth estimation accuracy but can also effectively improve the distance estimation accuracy when compared with several state-of-the-art supervised binaural distance estimation methods.

Enhanced Electrocatalytic Hydrogen Evolution Activity in Single-Atom Pt-Decorated VS2 Nanosheets.

Enhancing catalytic activity by decorating noble metals in catalysts provides an opportunity for promoting the electrocatalytic hydrogen evolution reaction (HER) application. However, there are few systematic studies on regulating the structures of noble metals in catalytic materials and investigating their influence on HER. Herein, Pt catalysts with different structures including single atoms (SAs), clusters, and nanoparticles well-controllably anchored on VS2 nanosheets through a cost-effective optothermal method are reported, and their HER performance is studied. The most efficient Pt-decorated VS2 catalyst (with both Pt SAs and clusters) delivers an overpotential of 77 mV at 10 mA cm-2, close to that of Pt/C (48 mV). However, the optimal mass activity of Pt (normalizing to Pt content) is obtained from only SA Pt-decorated VS2 (i.e., 22.88 A mgPt-1 at 200 mV) and is 12 times greater than that of the Pt/C (1.87 A mgPt-1), attributed to the greatly enhanced Pt utilization. Additionally, the theoretical simulations reveal that Pt SA decoration makes the adsorption free energy of H* closer to the thermoneutral value and improves the charge-transfer kinetics, significantly enhancing HER activity. This work offers a pathway to prepare the desired catalyst based on synergy of Pt structures and VS2 and reveals the intrinsic mechanism for enhancing catalytic activity, which is important for HER applications.