arXiv subjects

Rui Zhao

Publications and source records attributed to Rui Zhao.

316 records · Page 18Linked to original sources

arXiv2017

Acoustic-To-Word Model Without OOV

Recently, the acoustic-to-word model based on the Connectionist Temporal Classification (CTC) criterion was shown as a natural end-to-end model directly targeting words as output units. However, this type of word-based CTC model suffers from the out-of-vocabulary (OOV) issue as it can only model limited number of words in the output layer and maps all the remaining words into an OOV output node. Therefore, such word-based CTC model can only recognize the frequent words modeled by the network output nodes. It also cannot easily handle the hot-words which emerge after the model is trained. In this study, we improve the acoustic-to-word model with a hybrid CTC model which can predict both words and characters at the same time. With a shared-hidden-layer structure and modular design, the alignments of words generated from the word-based CTC and the character-based CTC are synchronized. Whenever the acoustic-to-word model emits an OOV token, we back off that OOV segment to the word output generated from the character-based CTC, hence solving the OOV or hot-words issue. Evaluated on a Microsoft Cortana voice assistant task, the proposed model can reduce the errors introduced by the OOV output token in the acoustic-to-word model by 30%.

cs.CL↗

arXiv2017

Polarization-multiplexed, dual-beam swept source optical coherence tomography angiography

A polarization-multiplexed, dual-beam setup is proposed to expand the field of view for a swept source optical coherence tomography angiography (OCTA) system. This method used a Wollaston prism to split sample path light into two orthogonal-polarized beams. This allowed two beams to shine on the cornea at an angle separation of ~ 14 degrees, which led to a separation of ~ 4.2 mm on the retina. A 3-mm glass plate was inserted into one of the beam paths to set a constant path length difference between the two polarized beams so the interferogram from the two beams are coded at different frequency bands. The resulting OCTA images from the two beams were coded with a depth separation of ~ 2 mm. 5x5 mm2 angiograms from the two beams were obtained simultaneously in 4 seconds. The two angiograms then were montaged to get a wider field of view (FOV) of ~ 5x9.2 mm2.

physics.med-ph↗

arXiv2017

Large-Scale Domain Adaptation via Teacher-Student Learning

High accuracy speech recognition requires a large amount of transcribed data for supervised training. In the absence of such data, domain adaptation of a well-trained acoustic model can be performed, but even here, high accuracy usually requires significant labeled data from the target domain. In this work, we propose an approach to domain adaptation that does not require transcriptions but instead uses a corpus of unlabeled parallel data, consisting of pairs of samples from the source domain of the well-trained model and the desired target domain. To perform adaptation, we employ teacher/student (T/S) learning, in which the posterior probabilities generated by the source-domain model can be used in lieu of labels to train the target-domain model. We evaluate the proposed approach in two scenarios, adapting a clean acoustic model to noisy speech and adapting an adults speech acoustic model to children speech. Significant improvements in accuracy are obtained, with reductions in word error rate of up to 44% over the original source model without the need for transcribed data in the target domain. Moreover, we show that increasing the amount of unlabeled data results in additional model robustness, which is particularly beneficial when using simulated training data in the target-domain.

cs.CL↗

arXiv2017

Automated Low-cost Terrestrial Laser Scanner for Measuring Diameters at Breast Height and Heights of Forest Trees

Terrestrial laser scanner is a kind of fast, high-precision data acquisition device, which had been more and more applied to the research areas of forest inventory. In this study, a kind of automated low-cost terrestrial laser scanner was designed and implemented based on a two-dimensional laser radar sensor SICK LMS-511 and a stepper motor. The new scanner was named as BEE, which can scan the forest trees in three dimension. The BEE scanner and its supporting software are specifically designed for forest inventory. The experiments have been performed by using the BEE scanner in an artificial ginkgo forest which was located in Haidian district of Beijing. Four square plots were selected to do the experiments. The BEE scanner scanned in the four plots and acquired the single scan data respectively. The DBH, tree height and tree position of trees in the four plots were estimated and analyzed. For comparison, the manual measured data was also collected in the four plots. The tree stem detection rate for all four plots was 92.75%; the root mean square error of the DBH estimation was 1.27cm; the root mean square error of the tree height estimation was 0.24m; the tree position estimation was in line with the actual position. Experimental results show that the BEE scanner can efficiently estimate the structure parameters of forest trees and has a good potential in practical application of forest inventory.

cs.CV↗

arXiv2017

A Nuclear-norm Model for Multi-Frame Super-Resolution Reconstruction from Video Clips

We propose a variational approach to obtain super-resolution images from multiple low-resolution frames extracted from video clips. First the displacement between the low-resolution frames and the reference frame are computed by an optical flow algorithm. Then a low-rank model is used to construct the reference frame in high-resolution by incorporating the information of the low-resolution frames. The model has two terms: a 2-norm data fidelity term and a nuclear-norm regularization term. Alternating direction method of multipliers is used to solve the model. Comparison of our methods with other models on synthetic and real video clips show that our resulting images are more accurate with less artifacts. It also provides much finer and discernable details.

cs.CV↗

arXiv2017

Generalized thermoelastic band structures of Rayleigh wave in one-dimensional phononic crystals

We investigate generalized band structures of Rayleigh wave in 1-D phononic crystals in the context of Green-Nagdhi thermoelastic (TE) theory. Solutions of the coupled equations included thermal field are obtained firstly. Then according to boundary conditions, phase velocity is derived. The transfer matrix methodology is adopted in order to formulate the band structures of generalized TE Rayleigh wave. It is demonstrated that the band structure of TE Rayleigh wave is comprised of elastic and thermal bands. For Aluminum-Epoxy phononic crystal, the thermoelasticity can influence the transmission ability of band gaps as well as narrow down the band gap width.

cond-mat.mtrl-sci↗

arXiv2016

Deep Learning and Its Applications to Machine Health Monitoring: A Survey

Since 2006, deep learning (DL) has become a rapidly growing research direction, redefining state-of-the-art performances in a wide range of areas such as object recognition, image segmentation, speech recognition and machine translation. In modern manufacturing systems, data-driven machine health monitoring is gaining in popularity due to the widespread deployment of low-cost sensors and their connection to the Internet. Meanwhile, deep learning provides useful tools for processing and analyzing these big machinery data. The main purpose of this paper is to review and summarize the emerging research work of deep learning on machine health monitoring. After the brief introduction of deep learning techniques, the applications of deep learning in machine health monitoring systems are reviewed mainly from the following aspects: Auto-encoder (AE) and its variants, Restricted Boltzmann Machines and its variants including Deep Belief Network (DBN) and Deep Boltzmann Machines (DBM), Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). Finally, some new trends of DL-based machine health monitoring methods are discussed.

cs.LG↗

arXiv2016

Zone-center phonons of bulk, few-layer, and monolayer 1T-TaS$_2$: Detection of the commensurate charge density wave phase through Raman scattering

We present first-principles calculations of the vibrational properties of the transition metal dichalcogenide 1T-TaS$_2$ for various thicknesses in the high-temperature (undistorted) phase and the low-temperature commensurate charge density wave (CDW) phase. We also present measurements of the Raman spectra for bulk, few-layer, and monolayer samples at temperatures well below that of the bulk transition to the commensurate phase. Through our calculations, we identify the low-frequency folded-back acoustic modes as a convenient signature of the commensurate CDW wave structure in vibrational spectra. In our measured Raman spectra, this signature is clearly evident in all of the samples, indicating that the commensurate phase remains the ground state as the material is thinned, even down to a single layer. This is in contrast to some previous studies which suggest a suppression of the commensurate CDW transition in thin flakes. We also use polarized Raman spectroscopy to probe c-axis orbital texture in the low-T phase, which has recently been suggested as playing a role in the metal-insulator transition that accompanies the structural transition to the commensurate CDW phase.

cond-mat.mtrl-sci↗

arXiv2014

Highly Efficient Forward and Backward Propagation of Convolutional Neural Networks for Pixelwise Classification

We present highly efficient algorithms for performing forward and backward propagation of Convolutional Neural Network (CNN) for pixelwise classification on images. For pixelwise classification tasks, such as image segmentation and object detection, surrounding image patches are fed into CNN for predicting the classes of centered pixels via forward propagation and for updating CNN parameters via backward propagation. However, forward and backward propagation was originally designed for whole-image classification. Directly applying it to pixelwise classification in a patch-by-patch scanning manner is extremely inefficient, because surrounding patches of pixels have large overlaps, which lead to a lot of redundant computation. The proposed algorithms eliminate all the redundant computation in convolution and pooling on images by introducing novel d-regularly sparse kernels. It generates exactly the same results as those by patch-by-patch scanning. Convolution and pooling operations with such kernels are able to continuously access memory and can run efficiently on GPUs. A fraction of patches of interest can be chosen from each training image for backward propagation by applying a mask to the error map at the last CNN layer. Its computation complexity is constant with respect to the number of patches sampled from the image. Experiments have shown that our proposed algorithms speed up commonly used patch-by-patch scanning over 1500 times in both forward and backward propagation. The speedup increases with the sizes of images and patches.

cs.CV↗

arXiv2014

Person Re-identification by Saliency Learning

Human eyes can recognize person identities based on small salient regions, i.e. human saliency is distinctive and reliable in pedestrian matching across disjoint camera views. However, such valuable information is often hidden when computing similarities of pedestrian images with existing approaches. Inspired by our user study result of human perception on human saliency, we propose a novel perspective for person re-identification based on learning human saliency and matching saliency distribution. The proposed saliency learning and matching framework consists of four steps: (1) To handle misalignment caused by drastic viewpoint change and pose variations, we apply adjacency constrained patch matching to build dense correspondence between image pairs. (2) We propose two alternative methods, i.e. K-Nearest Neighbors and One-class SVM, to estimate a saliency score for each image patch, through which distinctive features stand out without using identity labels in the training procedure. (3) saliency matching is proposed based on patch matching. Matching patches with inconsistent saliency brings penalty, and images of the same identity are recognized by minimizing the saliency matching cost. (4) Furthermore, saliency matching is tightly integrated with patch matching in a unified structural RankSVM learning framework. The effectiveness of our approach is validated on the VIPeR dataset and the CUHK01 dataset. Our approach outperforms the state-of-the-art person re-identification methods on both datasets.

cs.CV↗