词典驱动的联机手写日文病名识别研究

研究了一种有效的词典驱动的联机手写日文病名识别方法。病名词典以树结构存储,包含21 713个病名短语。在切分中,手写病名字符串通过分析相邻笔划之间的空间信息等特征被切分为原始的片段序列。连续的片段动态地合并为候选字符模式,不同的合并方式产生不同的候选字符序列,这样可构成一个切分候选网格。在识别过程中,结合病名词典匹配来限制候选字符模式的类别扩展,采用集束搜索策略来寻找到一条最优路径作为识别结果。用500个实际的手写病名样本做实验,平均每个病名的识别时间为0.87 s,识别正确率为83.16%。     

On-line handwritten Chinese character recognition via a fuzzy attribute representation

Chinese characters are constructed by strokes according to structural rules. Therefore, the geometric configurations of characters are important features for character recognition. In handwritten characters, stroke shapes and their spatial relations may vary to some extent. The attribute value of a structural identification is then a fuzzy quantity rather than a binary quantity. Recognizing these facts, we propose a fuzzy attribute representation (FAR) to describe the structural features of handwritten Chinese characters for an on-line Chinese character recognition (OLCCR) system. With a FAR. a fuzzy attribute graph for each handwritten character is created, and the character recognition process is thus transformed into a simple graph matching problem. This character representation and our proposed recognition method allow us to relax the constraints on stroke order and stroke connection. The graph model provides a generalized character representation that can easily incorporate newly added characters into an OLCCR system with an automatic learning capability. The fuzzy representation can describe the degree of structural deformation in handwritten characters. The character matching algorithm is designed to tolerate structural deformations to some extent. Therefore, even input characters with deformations can be recognized correctly once the reference dictionary of the recognition system has been trained using a few representative learning samples. Experimental results are provided to show the effectiveness of the proposed method.     

A robust method for coarse classifier construction from a large number of basic recognizers for on-line handwritten Chinese/Japanese character recognition

In this paper, a systematic method is described that constructs an efficient and a robust coarse classifier from a large number of basic recognizers obtained by different parameters of feature extraction, different discriminant methods or functions, etc. The architecture of the coarse classification is a sequential cascade of basic recognizers that reduces the candidates after each basic recognizer. A genetic algorithm determines the best cascade with the best speed and highest performance. The method was applied for on-line handwritten Chinese and Japanese character recognitions. We produced hundreds of basic recognizers with different classification costs and different classification accuracies by changing parameters of feature extraction and discriminant functions. From these basic recognizers, we obtained a rather simple two-stage cascade, resulting in the whole recognition time being reduced largely while maintaining classification and recognition rates.     

基于特征化字典的低秩表示人脸识别

针对现有低秩表示（LRR）算法中全局与局部人脸特征信息融合不足的问题,提出了一种新的人脸识别算法——基于特征化字典的低秩表示（LRR-CD）。首先,将每张人脸照片表示成一个个特征化字典的集合,然后同时最小化基于训练样本的低秩重构特征系数以及与之相对应的类内特征差异。为了获得高效且具有高判别性的人脸图像的特征块重构系数矩阵,提出了一种新的数学公式模型,通过同时求解训练样本中相对应的特征块以及对应的类内特征差异词典的低秩约束问题,尽可能完整地保留原始高维人脸图像中的全局和局部信息,尤其是局部类内差异特征。另外,由于对特征块中信息的充分挖掘,所提算法对于一般程度上的面部遮挡和光照等噪声影响具有良好的鲁棒性。在AR、CMU-PIE和Extended Yale B人脸数据库进行多项对比实验,由实验结果可知LRR-CD相较于对比的稀疏表示（SRC）、协从表示（CRC）、低秩表示正规切（LRR-NCUT）和低秩递归最小二乘（LRR-RLS）算法在平均识别率上有2.58~17.24个百分点的提高。实验结果表明LRR-CD性能优于与之对比的算法,可以更高效地用于人脸全局和局部特征信息的融合,且具有优良的识别率。     

解纠缠表示学习在跨年龄人脸识别中的应用

跨年龄人脸识别因其在现实生活中的广泛应用而成为人脸识别领域的热门话题.针对跨年龄人脸识别精度较低的问题,引入解纠缠表示学习,提出了一个基于生成对抗网络的解纠缠表示学习(IPDRL)网络来实现人脸图像的识别.该网络由编码器、生成器和鉴别器构成.编码器在对特征中的年龄变化进行解纠缠的同时,对人脸图像的身份信息进行编码,提取只利于身份鉴别的特征,实现身份特征和年龄特征的解纠缠;生成器根据输入的年龄特征生成对应的身份保持的年龄图像;鉴别器通过对抗学习和多任务学习实现年龄和身份的类分布预测.通过将解纠缠表示学习、对抗学习和多任务学习相结合的方法,很好地保留了人脸图像的身份信息,并使跨年龄人脸图像识别的精度得到了提高.     

Improving on-line handwritten recognition in interactive machine translation

On-line handwriting text recognition (HTR) could be used as a more natural way of interaction in many interactive applications. However, current HTR technology is far from developing error-free systems and, consequently, its use in many applications is limited. Despite this, there are many scenarios, as in the correction of the errors of fully-automatic systems using HTR in a post-editing step, in which the information from the specific task allows to constrain the search and therefore to improve the HTR accuracy. For example, in machine translation (MT), the on-line HTR system can also be used to correct translation errors. The HTR can take advantage of information from the translation problem such as the source sentence that is translated, the portion of the translated sentence that has been supervised by the human, or the translation error to be amended. Empirical experimentation suggests that this is a valuable information to improve the robustness of the on-line HTR system achieving remarkable results.     

A modal learning adaptive function neural network applied to handwritten digit recognition

A novel combination of the adaptive function neural network (ADFUNN) and on-line snap-drift learning is presented in this paper and applied to optical and pen-based recognition of handwritten digits [E. Alpaydin, F. Alimoglu for Optical Recognition of Handwritten Digits and E. Alpaydin, C. Kaynak for Pen-Based Recognition of Handwritten Digits http://www.ics.uci.edu/~mlearn/databases/optdigits/http://www.ics.uci.edu/~mlearn/databases/pendigits/]. Snap-drift [S.W. Lee, D. Palmer-Brown, C.M. Roadknight, Performance-guided neural network for rapidly self-organising active network management (Invited Paper), Journal of Neurocomputing, 61C, 2004, pp. 5-20] employs the complementary concepts of common (intersection) feature learning (called snap) and LVQ (drift towards the input patterns) learning, and is a fast, unsupervised method suitable for on-line learning and non-stationary environments where new patterns are continually introduced. ADFUNN [M. Kang, D. Palmer-Brown, An adaptive function neural network (ADFUNN) for phrase recognition, in: The International Joint Conference on Neural Networks (IJCNN05), Montréal, Canada, 2005, D. Palmer-Brown, M. Kang, ADFUNN: An adaptive function neural network, in: The 7th International Conference on Adaptive and Natural Computing Algorithms (ICANNGA05), Coimbra, Portugal, 2005] is based on a linear piecewise neuron activation function that is modified by a novel gradient descent supervised learning algorithm. It has recently been applied to the Iris dataset, and a natural language phrase recognition problem, exhibiting impressive generalisation classification ability with no hidden neurons. The unsupervised single layer snap-drift is effective in extracting distinct features from the complex cursive-letter datasets, and the supervised single layer ADFUNN is capable of solving linearly inseparable problems rapidly. In combination within one network (SADFUNN), these two methods are more powerful and yet simpler than MLPs, at least on this problem domain. We experiment on SADFUNN with two handwritten digits datasets problems from the UCI Machine Learning repository. The problems are learned rapidly and higher generalisation results are achieved than with a MLP.     

Robust face recognition via occlusion dictionary learning

Sparse representation based classification (SRC) has recently been proposed for robust face recognition. To deal with occlusion, SRC introduces an identity matrix as an occlusion dictionary on the assumption that the occlusion has sparse representation in this dictionary. However, the results show that SRC's use of this occlusion dictionary is not nearly as robust to large occlusion as it is to random pixel corruption. In addition, the identity matrix renders the expanded dictionary large, which results in expensive computation. In this paper, we present a novel method, namely structured sparse representation based classification (SSRC), for face recognition with occlusion. A novel structured dictionary learning method is proposed to learn an occlusion dictionary from the data instead of an identity matrix. Specifically, a mutual incoherence of dictionaries regularization term is incorporated into the dictionary learning objective function which encourages the occlusion dictionary to be as independent as possible of the training sample dictionary. So that the occlusion can then be sparsely represented by the linear combination of the atoms from the learned occlusion dictionary and effectively separated from the occluded face image. The classification can thus be efficiently carried out on the recovered non-occluded face images and the size of the expanded dictionary is also much smaller than that used in SRC. The extensive experiments demonstrate that the proposed method achieves better results than the existing sparse representation based face recognition methods, especially in dealing with large region contiguous occlusion and severe illumination variation, while the computational cost is much lower.     

Performance evaluation of pattern classifiers for handwritten character recognition

This paper describes a performance evaluation study in which some efficient classifiers are tested in handwritten digit recognition. The evaluated classifiers include a statistical classifier (modified quadratic discriminant function, MQDF), three neural classifiers, and an LVQ (learning vector quantization) classifier. They are efficient in that high accuracies can be achieved at moderate memory space and computation cost. The performance is measured in terms of classification accuracy, sensitivity to training sample size, ambiguity rejection, and outlier resistance. The outlier resistance of neural classifiers is enhanced by training with synthesized outlier data. The classifiers are tested on a large data set extracted from NIST SD19. As results, the test accuracies of the evaluated classifiers are comparable to or higher than those of the nearest neighbor (1-NN) rule and regularized discriminant analysis (RDA). It is shown that neural classifiers are more susceptible to small sample size than MQDF, although they yield higher accuracies on large sample size. As a neural classifier, the polynomial classifier (PC) gives the highest accuracy and performs best in ambiguity rejection. On the other hand, MQDF is superior in outlier rejection even though it is not trained with outlier data. The results indicate that pattern classifiers have complementary advantages and they should be appropriately combined to achieve higher performance. Received: July 18, 2001 / Accepted: September 28, 2001     

Position-width-pulse algorithm for recognition of handwritten characters

An algorithm for the recognition of handwritten characters based on the position-width-pulse method of recognition of curves is presented in the article. An algorithm for transformation of characters into curves is presented and a recognition procedure described. Numerical estimators of the proximity S _w of curves that graphically map the image of characters to be recognized relative to printed characters used as reference characters are calculated. A conclusion that assigns a recognized character to a corresponding reference character is arrived at on the basis of the minimum value of S _w with specified reliability.     

Query-expanded collaborative representation based classification with class-specific prototypes for object recognition

Linear representation based classifiers (LinearRCs) assume that a query image can be represented as a linear combination of dictionary atoms or prototypes with various priors (e.g., sparsity), which have achieved impressive results in face recognition. Recently, a few attempts have been made to deal with more general cases (e.g., multi-view or multi-pose objects, more generic objects, etc.) but with additional requirements. In this paper, we present a query-expanded collaborative representation based classifier with class-specific prototypes (QCRC_CP) from the general perspective. First, we expand a single query in a multi-resolution way to cover rich variations of object appearances, thereby generating a query set. We then condense the gallery images to a small amount of prototypical images by maximizing canonical correlation in a class-specific way, in which the implicit query-dependent data locality discards the outliers. Given the query set, we finally propose a multivariate LinearRC with collaborative prior to identify the query according to the rule of minimum normalized residual (MNR). Experiments on four object recognition datasets (FERET pose, Swedish leaf, Chars74K, and ETH-80) show that our method outperforms the state-of-the-art LinearRCs with performance increases at least 3.1%, 3.8%, 10.4% and 3.1% compared to other classifiers.     

Building a compact online MRF recognizer for large character set by structured dictionary representation and vector quantization technique

This paper describes a method for building a compact online Markov random field (MRF) recognizer for large handwritten Japanese character set using structured dictionary representation and vector quantization (VQ) technique. The method splits character patterns into radicals, whose models by MRF are shared by different character classes such that a character model is constructed from the constituent radical models. Many distinct radicals are shared by many character classes with the result that the storage space of model dictionary can be saved. Moreover, in order to further compress the parameters, VQ technique to cluster parameter sequences of the mean vectors and covariance matrixes for MRF unary features and binary features as well as the transition probabilities of each state into groups was employed. By sharing a common parameter sequence for each group, the dictionary of the MRF recognizer can be greatly compressed without recognition accuracy loss.     

The state of the art in Japanese online handwriting recognition compared to techniques in western handwriting recognition

This paper compares the current state of the art in online Japanese character recognition with techniques in western handwriting recognition. It discusses important developments in preprocessing, classification, and postprocessing for Japanese character recognition in recent years and relates them to the developments in western handwriting recognition. Comparing eastern and western handwriting recognition techniques allows learning from very different approaches and understanding the underlying common foundations of handwriting recognition. This is very important when it comes to developing compact modules for integrated systems supporting many writing systems capable of recognizing multilanguage documents.Received: January 12, 2002, Accepted: March 6, 2003, Published online: 4 July 2003     

Multilayer perceptrons combination applied to handwritten character recognition

Several methods of combination of Multilayer Perceptrons (MLPs) for handwritten character recognition are presented and discussed. Recognition tests have shown that cooperation of neural networks using different features vectors can reduce significantly the overall misclassification error rate. Additionally, the MLPs that are combined are the results of the experiments that were previously performed in order to optimize the recognition process when using a single MLP. So, all the combination methods that are proposed are very easy to carry out. The final recognition system consists of a cascade association of small MLPs, which allows minimization of the overall recognition time while retaining a high recognition rate. This system appears to be 2.5 times faster than the best of the individual MLPs, while offering a recognition rate of 99.8% on unconstrained digits extracted from the NIST 3 database.     

基于特征反馈的手写体汉字识别系统研究

针对传统基于开环的汉字识别系统不完全符合人类识字过程的问题,构建了一种具有反馈结构的仿人智能识别系统。该系统根据待识别汉字的多模态定性识别结果来选择最佳的首轮识别方案,在完成识别之后,提取广义识字误差对候选字进行可信度判断和反馈校正。设计了3种广义识字误差,通过对这3种广义识字误差的类型和数值进行定性与定量相结合的分析,建立了识别结果的可信度评价指标体系和反馈校正决策机制。仿真实验结果验证了方法的可行性。     

Online continual learning for human activity recognition

Sensor-based human activity recognition (HAR), with the ability to recognise human activities from wearable or embedded sensors, has been playing an important role in many applications including personal health monitoring, smart home, and manufacturing. The real-world, long-term deployment of these HAR systems drives a critical research question: how to evolve the HAR model automatically over time to accommodate changes in an environment or activity patterns. This paper presents an online continual learning (OCL) scenario for HAR, where sensor data arrives in a streaming manner which contains unlabelled samples from already learnt activities or new activities. We propose a technique, OCL-HAR, making a real-time prediction on the streaming sensor data while at the same time discovering and learning new activities. We have empirically evaluated OCL-HAR on four third-party, publicly available HAR datasets. Our results have shown that this OCL scenario is challenging to state-of-the-art continual learning techniques that have significantly underperformed. Our technique OCL-HAR has consistently outperformed them in all experiment setups, leading up to 0.17 and 0.23 improvements in micro and macro F1 scores.