Source author record

Yan Yin

Yan Yin appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

cond-mat.mtrl-sci physics.optics Computation and Language cond-mat.other eess.AS physics.acc-ph Sound

Catalog footprint

What is connected

5works

7topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2020arXiv

Attention-based Transducer for Online Speech Recognition

Recent studies reveal the potential of recurrent neural network transducer (RNN-T) for end-to-end (E2E) speech recognition. Among some most popular E2E systems including RNN-T, Attention Encoder-Decoder (AED), and Connectionist Temporal Classification (CTC), RNN-T has some clear advantages given that it supports streaming recognition and does not have frame-independency assumption. Although significant progresses have been made for RNN-T research, it is still facing performance challenges in terms of training speed and accuracy. We propose attention-based transducer with modification over RNN-T in two aspects. First, we introduce chunk-wise attention in the joint network. Second, self-attention is introduced in the encoder. Our proposed model outperforms RNN-T for both training speed and accuracy. For training, we achieves over 1.7x speedup. With 500 hours LAIX non-native English training data, attention-based transducer yields ~10.6% WER reduction over baseline RNN-T. Trained with full set of over 10K hours data, our final system achieves ~5.5% WER reduction over that trained with the best Kaldi TDNN-f recipe. After 8-bit weight quantization without WER degradation, RTF and latency drop to 0.34~0.36 and 268~409 milliseconds respectively on a single CPU core of a production server.

preprint2016arXiv

Longitudinal RF capture simulation and BPM signal estimation

In this paper, the theoretical aspects behind longitudinal RF capture are reviewed and the capture process is simulated via a program based on this theory. Four kinds of cases with different initial distribution and capture curve are considered, i.e. uniform distribution with adiabatic capture, uniform distribution with non-adiabatic capture, Gaussian distribution with adiabatic capture and Gaussian distribution with non-adiabatic capture. The simulation results are compared each other and discussed, and Gaussian distribution with adiabatic capture is demonstrated having a higher capture efficiency and leading to a shorter bunch length. In addition, the BPM induced signal is simulated with high input impendence, i.e. $1MΩ$, and low input impendence, i.e. $50Ω$, respectively. Finally, the BPM signal of Heavy Ion Medical Machine (HIMM) is estimated and compared with measured one, and a good agreement is achieved.

preprint2015arXiv

Mechanism and modulation of terahertz generation from a semimetal - graphite

Semi-metals might offer a stronger interaction and a better confinement for terahertz wave than semiconductors, while preserve tunability. Particularly, graphene-based materials are envisioned as terahertz modulators, filters and ultra-broadband sources. However, the understanding of terahertz generation from those materials is still not clear, thus limits us recognizing the potential and improving device performances. Graphite, the mother material of graphene and a typical bulk semi-metal, is a good system to study semi-metals and graphene-based materials. Here we experimentally modulate and maximize the terahertz signal from graphite surface, thus reveal the mechanism - surface field driving photon induced carriers into transient current to radiate terahertz wave. We also discuss the differences between graphite and semiconductors; particularly graphite shows no temperature dependency from room temperature to 80C. Above knowledge will help us understand terahertz generations, achieve maximum output and electric modulation, in semi-metal or graphene based devices.

preprint2014arXiv

Graphene, a material for high temperature devices; intrinsic carrier density, carrier drift velocity, and lattice energy

Heat has always been a killing matter for traditional semiconductor machines. The underlining physical reason is that the intrinsic carrier density of a device made from a traditional semiconductor material increases very fast with a rising temperature. Once reaching a temperature, the density surpasses the chemical doping or gating effect, any p-n junction or transistor made from the semiconductor will fail to function. Here, we measure the intrinsic Fermi level (|E_F|=2.93k_B*T) or intrinsic carrier density (n_in=3.87*10^6 cm^-2 K^-2*T^2), carrier drift velocity, and G mode phonon energy of graphene devices and their temperature dependencies up to 2400 K. Our results show intrinsic carrier density of graphene is an order of magnitude less sensitive to temperature than those of Si or Ge, and reveal the great potentials of graphene as a material for high temperature devices. We also observe a linear decline of saturation drift velocity with increasing temperature, and identify the temperature coefficients of the intrinsic G mode phonon energy. Above knowledge is vital in understanding the physical phenomena of graphene under high power or high temperature.

preprint2007arXiv

Screening of Excitons in Single, Suspended Carbon Nanotubes

Resonant Raman spectroscopy of single carbon nanotubes suspended across trenches displays red shifts of up to 30 meV of the electronic transition energies as a function of the surrounding dielectric environment. We develop a simple scaling relationship between the exciton binding energy and the external dielectric function and thus quantify the effect of screening. Our results imply that the underlying particle interaction energies change by hundreds of meV.