Extending noise robust structured support vector machines to larger vocabulary tasks

This paper describes a structured SVM framework suitable for noise-robust medium/large vocabulary speech recognition. Several theoretical and practical extensions to previous work on small vocabulary tasks are detailed. The joint feature space based on word models is extended to allow context-dependent triphone models to be used. By interpreting the structured SVM as a large margin log-linear model, illustrates that there is an implicit assumption that the prior of the discriminative parameter is a zero mean Gaussian. However, depending on the definition of likelihood feature space, a non-zero prior may be more appropriate. A general Gaussian prior is incorporated into the large margin training criterion in a form that allows the cutting plan algorithm to be directly applied. To further speed up the training process, 1-slack algorithm, caching competing hypothesis and parallelization strategies are also proposed. The performance of structured SVMs is evaluated on noise corrupted medium vocabulary speech recognition task: AURORA 4. © 2011 IEEE.

Discriminative spoken language understanding using word confusion networks

Current commercial dialogue systems typically use hand-crafted grammars for Spoken Language Understanding (SLU) operating on the top one or two hypotheses output by the speech recogniser. These systems are expensive to develop and they suffer from significant degradation in performance when faced with recognition errors. This paper presents a robust method for SLU based on features extracted from the full posterior distribution of recognition hypotheses encoded in the form of word confusion networks. Following [1], the system uses SVM classifiers operating on n-gram features, trained on unaligned input/output pairs. Performance is evaluated on both an off-line corpus and on-line in a live user trial. It is shown that a statistical discriminative approach to SLU operating on the full posterior ASR output distribution can substantially improve performance both in terms of accuracy and overall dialogue reward. Furthermore, additional gains can be obtained by incorporating features from the previous system output. © 2012 IEEE.

Influence of specific HSP70 domains on fibril formation of the yeast prion protein Ure2.

Ure2p is the protein determinant of the Saccharomyces cerevisiae prion state [URE3]. Constitutive overexpression of the HSP70 family member SSA1 cures cells of [URE3]. Here, we show that Ssa1p increases the lag time of Ure2p fibril formation in vitro in the presence or absence of nucleotide. The presence of the HSP40 co-chaperone Ydj1p has an additive effect on the inhibition of Ure2p fibril formation, whereas the Ydj1p H34Q mutant shows reduced inhibition alone and in combination with Ssa1p. In order to investigate the structural basis of these effects, we constructed and tested an Ssa1p mutant lacking the ATPase domain, as well as a series of C-terminal truncation mutants. The results indicate that Ssa1p can bind to Ure2p and delay fibril formation even in the absence of the ATPase domain, but interaction of Ure2p with the substrate-binding domain is strongly influenced by the C-terminal lid region. Dynamic light scattering, quartz crystal microbalance assays, pull-down assays and kinetic analysis indicate that Ssa1p interacts with both native Ure2p and fibril seeds, and reduces the rate of Ure2p fibril elongation in a concentration-dependent manner. These results provide new insights into the structural and mechanistic basis for inhibition of Ure2p fibril formation by Ssa1p and Ydj1p.

Influence of specific HSP70 domains on fibril formation of the yeast prion protein Ure2.

Ure2p is the protein determinant of the Saccharomyces cerevisiae prion state [URE3]. Constitutive overexpression of the HSP70 family member SSA1 cures cells of [URE3]. Here, we show that Ssa1p increases the lag time of Ure2p fibril formation in vitro in the presence or absence of nucleotide. The presence of the HSP40 co-chaperone Ydj1p has an additive effect on the inhibition of Ure2p fibril formation, whereas the Ydj1p H34Q mutant shows reduced inhibition alone and in combination with Ssa1p. In order to investigate the structural basis of these effects, we constructed and tested an Ssa1p mutant lacking the ATPase domain, as well as a series of C-terminal truncation mutants. The results indicate that Ssa1p can bind to Ure2p and delay fibril formation even in the absence of the ATPase domain, but interaction of Ure2p with the substrate-binding domain is strongly influenced by the C-terminal lid region. Dynamic light scattering, quartz crystal microbalance assays, pull-down assays and kinetic analysis indicate that Ssa1p interacts with both native Ure2p and fibril seeds, and reduces the rate of Ure2p fibril elongation in a concentration-dependent manner. These results provide new insights into the structural and mechanistic basis for inhibition of Ure2p fibril formation by Ssa1p and Ydj1p.

Validation of a method of estimating age, modelling growth, and describing the age composition of Coilia mystus from the Yangtze Estuary, China

A collection of 577 Coilia mystus was made during April 2006 and 2007 from China's Yangtze Estuary to estimate the age structure and growth patterns of the population. Examination of sectioned sagittal otoliths revealed a periodic straight/curved growth pattern. The straight zone was from April to November, and the curved zone from October to May, indicating annual periodicity. Annual periodicity was also verified by margin zone analysis. The shift from a curved-zone to the next straight-zone stanza was defined as an annulus. The fish from which the otoliths were taken were 0-5 years old. The von Bertalanffy growth function was fitted to standard length (LS)-at-age data as L-S = 215.16 (1 - e(-0.53(t+0.30))) (n = 577, r(2) = 0.81, p < 0.05). The mature females included five age classes, ages 1 and 2 accounting for 74.3% of the population. The mature males included fish aged 1 and 2, those at age 1 accounting for 86.4% of the population. Mean length was smaller, and annual growth less, for mature males than for females of comparable age. The study demonstrated that the Yangtze population of C. mystus consists of more age classes than previously thought and that the age structure of the population needs to be considered in management decisions.

On-chip tunable micro-ring resonator based on digital microfluidics platform

We demonstrate the tunability of a silicon nitride micro-resonator using the concept of Digital Microfluidics. Our system allows driving micro-droplets on-chip, enabling the control of the effective refractive index at the vicinity of the resonator. © 2010 OSA/FiO/LS 2010.

长波长InAs/GaAs 量子点的MOCVD 生长及激光器制作

Submitted by zhangdi (zhangdi@red.semi.ac.cn) on 2009-04-13T11:45:31Z

High repetition rate Q-switched microchip Nd:YVO4 laser with pulse duration as short as 1.1 ns

We have demonstrated a compact and an efficient passively Q-switched microchip Nd:YVO4 laser by using a composite semiconductor absorber as well as an output coupler. The composite semiconductor absorber was composed of an LT (low-temperature grown) In0.25Ga0.75As absorber and a pure GaAs absorber. To our knowledge, it was the first demonstration of the special absorber for Q-switching operation of microchip lasers. Laser pulses with durations of 1.1 ns were generated with a 350 mu m thick laser crystal and the repetition rate of the pulses was as high as 4.6 MHz. The average output power was 120 mW at the pump power of 700 mW. Pulse duration can be varied from 1.1 to 15.7 ns by changing the cavity length from 0.45 to 5 mm. Pulses with duration of 1.67 and 2.41 ns were also obtained with a 0.7 mm, thick laser crystal and a 1 mm thick laser crystal, respectively. (C) 2007 Elsevier GmbH. All rights reserved.

Self-pulsation in a two-section DFB laser with a varied ridge width

A 1.55 mu m InGaAsP-InP two-section DFB laser with a variable ridge width has been fabricated. Self-pulsations with frequencies around 3 GHz and 40 GHz are observed. The pulsation mechanisms related to the two frequencies are discussed and the tunability of generated self-pulsations is studied.

Passive stabilization of a passively mode-locked Nd : GdVO4 laser by inverse saturable absorption

A diode-pumped Nd:GdVO4 laser mode-locked by a semiconductor saturable absorber and output coupler (SESAOC) is passively stabilized to suppress Q-switched mode-locking. A phase mismatched 131130 second-harmonic generation (SHG) crystal is used for passive stabilization. The continuous wave mode-locking (CWML) threshold is reduced and the pulse width is compressed. The pulse width is 6.5 ps as measured at the repetition rate of 128 MHz. (c) 2007 Elsevier B.V. All rights reserved.

Application of sampled grating to control the lasing wavelength in complex-coupled DFB laser

This paper reports that the complex-coupled distributed feedback laser with the sampled grating has been designed and fabricated. The +1st order reflection of the sampled grating is utilized for laser single mode operation, which is 1.5387 mu m in the experiment. The typical threshold current of the device is 30 mA, and the optical output power is about 10 mW at the injection current of 100 mA.

Study on Closed-set Speaker Identification Based on Biomimetic Pattern Recognition

In this paper, a new classifier of speaker identification has been proposed, which is based on Biomimetic pattern recognition (BPR). Distinguished from traditional speaker recognition methods, such as DWT, HMM, GMM, SVM and so on, the proposed classifier is constructed by some finite sub-space which is reasonable covering of the points in high dimensional space according to distributing characteristic of speech feature points. It has been used in the system of speaker identification. Experiment results show that better effect could be obtained especially with lesser samples. Furthermore, the proposed classifier employs a much simpler modeling structure as compared to the GMM. In addition, the basic idea "cognition" of Biomimetic pattern recognition (BPR) results in no requirement of retraining the old system for enrolling new speakers.

Control of the lasing wavelength by a sampled grating in a complex-coupled DFB laser

The lasing wavelength of a complex-coupled DFB laser is controlled by a sampled grating. The key concepts of the approach are to utilize the -1st order (negative first order) reflection of a sampled grating for laser single mode operation, and use conventional holographic exposure combined with the usual photolithography to fabricate the sampled grating. The typical threshold current of the sampled grating based DFB laser is 32 mA, and the optical output is about 10 mW at an injected current of 100 mA. The lasing wavelength of the device is 1.5356 mu m, which is the -1st order wavelength of the sampled grating.

Q-switched and mode-locked diode-pumped Nd : YVO4 laser with an intracavity composite semiconductor saturable absorber

We have demonstrated a passively Q-switched and mode-locked Nd:YVO4 laser with an intracavity composite semiconductor saturable absorber (ICSSA). Stable Q-switched and mode-locked pulses with Q-switched envelope pulse duration of 180 ns and pulse repetition rate of 72KHz have been obtained. The maximum average output power was 1.45W at 8W incident pump power. The repetition rate of the mode-locked pulses inside the Q-switched envelope was 154 MHz. Experimental results revealed that this ICSSA was suitable for Q-switched and mode-locked solid-state lasers. (c) 2007 Elsevier Ltd. All rights reserved.