Search

Surely no popular genre is as fascinated with origins as the superhero story, which almost invariably features an ordinary being suddenly and unexpectedly transformed into something extraordinary –...

Joe Simon and Jack Kirby’s Captain America famously burst into view when the debut issue of Captain America Comics appeared on newsstands during the final days of 1940 (despite the cover date of Ma...

We explore a variety of neural networks configurations for one- and two-channel spectrogram-mask-based speech enhancement. Our best model improves on previous state-of-the-art performance on the CHiME2 speech enhancement task by 0.4 decibels in signal-to-distortion ratio (SDR). We examine trade-offs such as non-causal look-ahead, computation, and parameter count versus enhancement performance and find that zero-look-ahead models can achieve, on average, within 0.03 dB SDR of our best bidirectional model. Further, we find that 200 milliseconds of look-ahead is sufficient to achieve equivalent performance to our best bidirectional model.

In recent years, deep networks have led to dramatic improvements in speech enhancement by framing it as a data-driven pattern recognition problem. In many modern enhancement systems, large amounts of data are used to train a deep network to estimate masks for complex-valued short-time Fourier transforms (STFTs) to suppress noise and preserve speech. However, current masking approaches often neglect two important constraints: STFT consistency and mixture consistency. Without STFT consistency, the system's output is not necessarily the STFT of a time-domain signal, and without mixture consistency, the sum of the estimated sources does not necessarily equal the input mixture. Furthermore, the only previous approaches that apply mixture consistency use real-valued masks; mixture consistency has been ignored for complex-valued masks. In this paper, we show that STFT consistency and mixture consistency can be jointly imposed by adding simple differentiable projection layers to the enhancement network. These layers are compatible with real or complex-valued masks. Using both of these constraints with complex-valued masks provides a 0.7 dB increase in scale-invariant signal-to-distortion ratio (SI-SDR) on a large dataset of speech corrupted by a wide variety of nonstationary noise across a range of input SNRs.

COSMA: Coherent Optics Sensors for Medical Application is an European Marie Curie Project running from 2012 to March 2016, with the participation of 10 teams from Armenia, Bulgaria, India, Israel, Italy, Poland, Russia, UK, USA. The main objective was to focus theoretical and experimental research on biomagnetism phenomena, with the specific aim to develop all-optical sensors dedicated to their detection and suitable for applications in clinical diagnostics. The paper presents some of the most recent results obtained during the exchange visits of the involved scientists, after an introduction about the phenomenon which is the pillar of this kind of research and of many other new fields in laser spectroscopy, atomic physics, and quantum optics: the dark resonance.

One of the pleasures of exploring the first blossoming of comic books in America lies in discovering how very strange some of the fruit was. Certainly, any reader approaching the Wonder Woman stori...

We use nuclear magnetic resonance spectroscopy of alkali metals sealed in glass vapor cells to perform in situ identification of chemical contaminants. The alkali Knight shift varies with the concentration of the impurity, which in turn varies with temperature as the alloy composition changes along the liquidus curve. Intentional addition of a known impurity validates this approach and reveals that sodium is often an intrinsic contaminant in cells filled with distilled, high-purity rubidium or cesium. Measurements of the Knight shift of the binary Rb–Na alloy confirm prior measurements of the shift's linear dependence on Na concentration, but similar measurements for the Cs–Na system demonstrate an unexpected nonlinear dependence of the Knight shift on the molar ratio. This non-destructive approach allows monitoring and quantification of ongoing chemical processes within the kind of vapor cells which form the basis for precise sensors and atomic frequency standards.