Topic-based language model has attracted much attention as the propounding of semantic retrieval in recent years. Especially for the ASR text with errors, the topic representation is more reasonable than the exact term representation. Among these models, Latent Dirichlet Allocation(LDA) has been noted for its ability to discover the latent topic structure, and is broadly applied in many text-related tasks. But up to now its application in information retrieval(IR) is still limited to be a supplement to the standard document models, and furthermore, it has been pointed out that directly employing the basic LDA model will hurt retrieval performance. In this paper, we propose a lexicon-guided two-level LDA retrieval framework. It uses the HowNet to guide the first-level LDA model's parameter estimation, and further construct the second-level LDA models based on the first-level's inference results. We use TRECID 2005 ASR collection to evaluate it, and compare it with the vector space model(VSM) and latent semantic Indexing(LSI). Our experiments show the proposed method is very competitive.

/pdf/lda-based-retrieval-framework-for-semantic-news-video-1l8b7tozus.pdf

LDA-Based Retrieval Framework for Semantic News Video Retrieval

Classification refers to associating a given fingerprint to one of the existing classes already recognized in the literature. A search over all the records in the database takes a long time, so the aim is to reduce the size of the search space by choosing an appropriate subset of database for search. Classifying a fingerprint images is a very difficult pattern recognition problem, due to the small interclass variability, the large intraclass variability. This paper presents a sequence flow diagram which will help in developing the clarity on designing algorithm for classification based on various parameters extracted from the fingerprint image. It discusses in brief the ways in which the parameters are extracted from the image. Existing fingerprint classification approaches are based on these parameters as input for classifying the image. Parameters like orientation map, singular points, spurious singular points, ridge flow and hybrid feature are discussed in the paper.

Review of Parameters of Fingerprint Classification Methods Based on Algorithmic Flow

In this paper, a supervised multi-class classification approach called Adaptive Selection of Information Components (ASIC) is presented. ASIC has the facilities to (i) handle both numerical and nominal features in a data set, (ii) pre-process the training data set to accentuate the spatial differences among the classes in the training data set to reduce further computational load requirements, and (iii) conduct supervised classification with the C-RSPM (Collateral Representative Subspace Projection Modeling) approach. Experimental results on a variety of data sets have shown that the proposed ASIC approach outperforms other well-known supervised classification methods such as C4.5, KNN, SVM, MLP, BN, RF, Logistic, and C-RSPM, with higher classification accuracy, lower training and classification times, and reduced memory storage and processing power requirements.

http://users.cis.fiu.edu/~chens/PDF/ICSC07.pdf

ASIC: Supervised Multi-class Classification using Adaptive Selection of Information Components

Classification refers to assigning a given fingerprint to one of the existing classes already recognized in the literature. A search over all the records in the database takes a long time, so the goal is to reduce the size of the search space by choosing an appropriate subset of database for search. Classifying a fingerprint images is a very difficult pattern recognition problem, due to the minimal interclass variability and maximal intraclass variability. This paper presents a sequence flow diagram which will help in developing the clarity on designing algorithm for classification based on various parameters extracted from the fingerprint image. It discusses in brief the ways in which the parameters are extracted from the image. Existing fingerprint classification approaches are based on these parameters as input for classifying the image. Parameters like orientation map, singular points, spurious singular points, ridge flow, transforms and hybrid feature are discussed in the paper.

/pdf/survey-on-parameters-of-fingerprint-classification-methods-4glq787irz.pdf

Survey on Parameters of Fingerprint Classification Methods Based On Algorithmic Flow

It is very important to detect the reference point effectively and accurately, especially in nonminutiae based fingerprint matching and fingerprint classification. In this paper, an effective algorithm based on block level for fingerprint reference point detection is proposed to simplify detection process and improve accuracy of the position of the point. The method is simple in the preprocessing and easy to implement. We used the Poincare Index on the block level, which is combined with the adaptive smoothing for getting a better orientation map and the directional consistency factor with the purpose of choosing the correct block. The proposed algorithm has been tested on fvc2002 and fvc2004 database. Experimental results show it is effective and practical.

An effective algorithm for fingerprint reference point detection

In this paper an effective fingerprint classification method based on Multiple Discriminant Analysis (MDA) is presented. The typology information is used to classify fingerprints. We also describe a feature extraction method based on Gabor filters. The effectiveness of our method is based on matters, that a reference point is searched from the region of interest and after finding the reference point, we crop the smaller image below the reference point. Then every operation, like feature calculation, is done only to this smaller image. The method was tested using artificially generated fingerprint database. With a uniform fingerprint distribution our classifier works an accuracy of 95.8% for the five-class problem.

https://ieeexplore.ieee.org/iel5/8534/27060/01201938.pdf

M. Tico

Papers

Fingerprint classification based on multiple discriminant analysis