In this paper, we propose multi-band MelGAN, a much faster waveform generation model targeting to high-quality text-to-speech. Specifically, we improve the original MelGAN by the following aspects. First, we increase the receptive field of the generator, which is proven to be beneficial to speech generation. Second, we substitute the feature matching loss with the multi-resolution STFT loss to better measure the difference between fake and real speech. Together with pre-training, this improvement leads to both better quality and better training stability. More importantly, we extend MelGAN with multiband processing: the generator takes mel-spectrograms as input and produces sub-band signals which are subsequently summed back to full-band signals as discriminator input. The proposed multi-band MelGAN has achieved high MOS of 4.34 and 4.22 in waveform generation and TTS, respectively. With only 1.91M parameters, our model effectively reduces the total computational complexity of the original MelGAN from 5.85 to 0.95 GFLOPS. Our Pytorch implementation, which will be open-resourced shortly, can achieve a real-time factor of 0.03 on CPU without hardware specific optimization.
Grazing-incidence small-angle X-ray scattering (GISAXS) is increasingly used for the metrology of substrate-supported nanoscale features and nanostructured films. In the case of line gratings, where long objects are arranged with a nanoscale periodicity perpendicular to the beam, a series of characteristic spots of high-intensity (grating truncation rods, GTRs) are recorded on a twodimensional detector. The intensity of the GTRs is modulated by the threedimensional shape and arrangement of the lines. Previous studies aimed to extract an average cross-sectional profile of the gratings, attributing intensity loss at GTRs to sample imperfections. Such imperfections are just as important as the average shape when employing soft polymer gratings which display significant line-edge roughness. Herein are reported a series of GISAXS measurements of polymer line gratings over a range of incident angles. Both an average shape and fluctuations contributing to the intensity in between the GTRs are extracted. The results are critically compared with atomic force microscopy (AFM) measurements, and it is found that the two methods are in good agreement if appropriate corrections for scattering from the substrate (GISAXS) and contributions from the probe shape (AFM) are accounted for.
A novel 16-channel fixed frequency Doppler backward scattering (DBS) reflectometer system has been developed on the HL-2A tokamak. This system is based on the filter-based feedback loop microwave source (FFLMS) technique, which has lower phase noise and lower power variation compared with present tunable frequency generation and comb frequency array generation techniques [J. C. Hillesheim et al. Rev. Sci. Instrum. 80, 083507 (2009) and W. A. Peebles et al. Rev. Sci. Instrum. 81, 10D902 (2010)]. The 16-channel DBS system is comprised of four × four-frequency microwave transmitters and direct quadrature demodulation receivers. The working frequencies are 17-24 GHz and 31-38 GHz with the frequency interval of 1 GHz. They are designed to measure the localized intermediate wave-number (kρ ∼ 1-2, k ∼ 2-9 cm) density fluctuations and the poloidal rotation velocity profile of turbulence. The details of the system design and laboratory tests are presented. Preliminary results of Doppler spectra measured by the multi-channel DBS reflectometer systems are obtained. The plasma rotation and turbulence distribution during supersonic molecular beam injection are analyzed.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.