Keyword-guided word spotting in historical printed documents using synthetic data and user feedback

In this paper, we propose a novel technique for word spotting in historical printed documents combining synthetic data and user feedback. Our aim is to search for keywords typed by the user in a large collection of digitized printed historical documents. The proposed method consists of the following stages: (1) creation of synthetic image words; (2) word segmentation using dynamic parameters; (3) efficient feature extraction for each word image and (4) a retrieval procedure that is optimized by user feedback. Experimental results prove the efficiency of the proposed approach.     

1000kN模具研配压力机机架有限元分析及优化设计

利用Cosmos Works对模具研配压力机的机架进行了有限元静力分析,求出了机架的应力和变形分布情况。根据静力学分析结果,对整个机架进行结构参数优化设计,使机架重量减少了9．28％,获得了良好的结构参数。对模具研配压力机框架式机架的设计提供了新的思路和方法。     

Identification of Codon 146 KRAS Variants in Isolated Epidermal Nevus and Multiple Lesions in Oculoectodermal Syndrome: Confirmation of the Phenotypic Continuum of Mosaic RASopathies

Mosaic RASopathies are a molecularly heterogeneous group of (neuro)cutaneous syndromes with high phenotypical variability. Postzygotic variants in KRAS have been described in oculoectodermal syndrome (OES), encephalocraniocutaneous lipomatosis (ECCL) and epidermal nevus syndrome (ENS). This study confirms the continuum of mosaic neurocutaneous RASopathies showing codon 146 KRAS variants in an individual with OES and, for the first time, in an individual with (isolated) epidermal nevus. The presence of a nevus psiloliparus in individuals with OES indicates that this finding is not specific for ECCL and highlights the phenotypical overlap between ECCL and OES. The presence of the somatic KRAS variant in the nevus psiloliparus resolves the underlying molecular etiology of this fatty-tissue nevus. In addition, this finding refutes the theory of non-allelic twin-spotting as an underlying hypothesis to explain the concurrent presence of two different mosaicisms in one individual. The identification of codon 146 KRAS variants in isolated epidermal nevus introduces a new hot spot for this condition, which is useful for increasing molecular genetic testing using targeted gene sequencing panels.     

人脸特征的定位与提取

人脸的特征定位与特征提取的好坏对于人脸图象识别效果有直接的影响．本文根据人脸图象的灰度特性以及特征的几何形状提出了一种简便有效的特征定位和特征提取的方法。本文所提出的方法能够比较准确而快速的提取较多的特征，因而能够适用于在大量人像中进行特征提取和识别。     

开孔导流管喷动床烟气脱硫技术实验研究

研究了一种新型的半干法脱硫开孔导流管喷动床。在相同静床层高度下,研究了其最小喷动速度、最大喷动压降、喷动高度;在相同钙硫摩尔比、入口SO2质量浓度、进气温度、绝热饱和温差条件下,以消石灰为脱硫剂,研究了其脱硫性能。实验结果表明,与传统柱锥型喷动床相比,开孔导流管喷动床具有更好的流体力学性能和脱硫效率。实验条件下最适宜的操作条件为:钙硫摩尔比为0.9—2.5,进气温度为120—130℃,绝热饱和温差为10℃。     

一种基于互补声学模型的多系统融合语音关键词检测方法

采用一种基于互补声学模型的多系统融合方法来获得高性能的语音关键词检测系统: 1)在基线系统的基础上, 使用不同的音素集进行声学建模, 并引入基于神经网络的声学建模方法, 获得另外两套具有建模差异性的声学系统; 2)在多套关键词检测系统的基础上, 通过选择有效的系统融合准则, 将多个系统的输出进行整合, 获得更好的语音关键词检测结果. 该方法充分利用了差异性声学建模系统之间的互补性, 在不增加训练数据的情况下, 显著地提升了最终系统的性能. 和基线系统相比, 该方法在2005年国家863电话语音关键词检测技术评测集上, 在等错误率(Equal error rate, EER)指标下, 获得相对21.6%的显著性能提升.     

一种轧机主轴定位系统在武钢CSP厂的应用

自动换辊功能是精轧机最重要的控制功能之一,主轴定位是实现自动换辊的前提条件。主轴等传动设备的位置控制实际上是电机速度控制的问题,结合武钢CSP厂的实际,讨论了主轴位置控制的方法,并介绍了一种主轴定位系统。     

A Two-step Keyword Spotting Method Using Fuzzy Search Algorithm

For many practical applications, input signal is a spontaneous conversation while the acoustic model was trained with read speech because of data availability. The performance of Keyword spotting system will degrade significantly. To solve this problem, we presented a two- step keyword spotting strategy. A fuzzy search algorithm was adopted to improve the retrieval performance. We generated confusion matrix from confusion network. It gives more freedom in the representation so as to alleviate the effect phoneme mlsrecognition. Furthermore a hybrid confidence measure was applied to reject false accepts. Experiments show that the proposed algorithms significantly reduce Equal error rate （EER） on the telephone conversational task.     

Hierarchical deep belief networks based point process model for keywords spotting in continuous speech

Point process model keyword spotting (KWS) system has attracted considerable attentions in the areas of keyword spotting by its capacity that can generalize from a relatively small numbers of training examples. But unfortunately, the accuracy level of the point process model is not comparable with the state‐of‐the‐art KWS systems because of the poor modeling capacity of the phoneme detector, which are based on Gaussian Mixture Models. In this paper, focus on improving the performance of detector in point process model, we propose an enhanced version of point process model, which is based on hierarchical deep belief networks (DBNs). Hierarchical DBNs are used as the phoneme detector in this system, and they combine the advantages of both the DBN and the hierarchical architecture for capturing complex statistical patterns in speech while overcoming the inherent flaws of conventional hidden Markov models and multilayer layer perceptron. Experiments results on TIMIT database show that the proposed method can yield 2% improvement. Furthermore, in the case when training examples are extremely limited, it can achieve better results over state‐of‐the‐art KWS systems. Copyright © 2013 John Wiley & Sons, Ltd.     

Simultaneous spotting of signs and fingerspellings based on hierarchical conditional random fields and boostmap embeddings

A sign language consists of two types of action; signs and fingerspellings. Signs are dynamic gestures discriminated by continuous hand motions and hand configurations, while fingerspellings are a combination of continuous hand configurations. Sign language spotting is the task of detection and recognition of signs and fingerspellings in a signed utterance. The internal structures of signs and fingerspellings differ significantly. Therefore, it is difficult to spot signs and fingerspellings simultaneously. In this paper, a novel method for spotting signs and fingerspellings is proposed. It can distinguish signs, fingerspellings and non-sign patterns, and is robust to the various sizes, scales and rotations of the signer's hand. This is achieved through a hierarchical framework consisting of three steps: (1) Candidate segments of signs and fingerspellings are discriminated using a two-layer conditional random field (CRF). (2) Hand shapes of segmented signs and fingerspellings are verified using BoostMap embeddings. (3) The motions of fingerspellings are verified in order to distinguish those which have similar hand shapes and different hand motions. Experiments demonstrate that the proposed method can spot signs and fingerspellings from utterance data at rates of 83% and 78%, respectively.