• Proceedings of the IEEK Conference
• /
• 2002.07c
• /
• pp.1670-1673
• /
• 2002

For pattern recognition on high-dimensional data, such as images, the dimensionality reduction as a preprocessing is effective. By dimensionality reduction, we can (1) reduce storage capacity or amount of calculation, and (2) avoid "the curse of dimensionality" and improve classification performance. Popular tools for dimensionality reduction are Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and Independent Component Analysis (ICA) recently. Among them, only LDA takes the class labels into consideration. Nevertheless, it, has been reported that, the classification performance with ICA is better than that with LDA because LDA has restriction on the number of dimensions after reduction. To overcome this dilemma, we propose a new dimensionality reduction technique based on an information theoretic measure for difference of distribution. It takes the class labels into consideration and still it does not, have restriction on number of dimensions after reduction. Improvement of classification performance has been confirmed experimentally.

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.1019-1022
• /
• 1999

In this paper, we analysis the effect of a dictionary in a handwritten Hangul word recognition problem in terms of its size and the length of the words in it. With our experimental results, we can account for the word recognition rate depending not only on character recognition performance, but also much on the amount of the information that the dictionary contains, as well as the reduction rate of a dictionary.

• Proceedings of the KSPS conference
• /
• 2003.05a
• /
• pp.123-126
• /
• 2003

We discuss how to reduce the number of inverse matrix and its dimensions requested in MLLR framework for speaker adaptation. To find a smaller set of variables with less redundancy, we employ PCA(principal component analysis) and ICA(independent component analysis) that would give as good a representation as possible. The amount of additional computation when PCA or ICA is applied is as small as it can be disregarded. The dimension of HMM parameters is reduced to about 1/3 ~ 2/7 dimensions of SI(speaker independent) model parameter with which speech recognition system represents word recognition rate as much as ordinary MLLR framework. If dimension of SI model parameter is n, the amount of computation of inverse matrix in MLLR is proportioned to O($n^4$). So, compared with ordinary MLLR, the amount of total computation requested in speaker adaptation is reduced to about 1/80~1/150.

• International journal of advanced smart convergence
• /
• v.7 no.4
• /
• pp.40-49
• /
• 2018

We propose a more stable robust recognition algorithm which detects faces reliably even in cases where there are changes in lighting and angle of view, as well it satisfies efficiency in calculation and detection performance. We propose detects the face area alone after normalization through pre-processing and obtains a feature vector using (PCA). The feature vector is applied to LDA and using Euclidean distance of intra-class variance and inter class variance in the 2nd dimension, the final analysis and matching is performed. Experimental results show that the proposed method has a wider distribution when the input image is rotated $45^{\circ}$ left / right. We can improve the recognition rate by applying this feature value to a single algorithm and complex algorithm, and it is possible to recognize in real time because it does not require much calculation amount due to dimensional reduction.

• The Journal of the Acoustical Society of Korea
• /
• v.10 no.3
• /
• pp.31-36
• /
• 1991

We propose an efficient pre-classification algorithm extracting candidate words to reduce the recognition time in a large-vocabulary recognition system and also propose the use of spectral and temporal smoothing of the observation probability to improve its classification performance. The proposed algorithm computes the coarse likelihood score for each word in a lexicon using the observation probabilities of speech spectra and duration information of recognition units. With the proposed approach we could reduce the computational amount by 74% with slight degradation of recognition accuracy in 1160-word recognition system based on the phoneme-level HMM. Also, we observed that the proposed coarse likelihood score computation algorithm is a good estimator of the likelihood score computed by the Viterbi algorithm.

• Proceedings of the IEEK Conference
• /
• 2000.11d
• /
• pp.199-202
• /
• 2000

Speaking Rate has variety depends on the situation and habit of speakers. It has been many studied about speaking rate In speaker recognition. The study of speaking rate in speech recognition is one of considerable matter when It is recognized the speakers and it is measured by many speech data base and complicate estimation for accuracy. In this paper, conventional vocoder process the speech signal when encoding and transmitting without regard to speaking rate so in order to apply the speaking rate for vocoder It should be considered the simpler algorithm and less computation amount than the conventional method of speaking rate used In speech recognition. We proposed the speaking rate algorithm which is used the simple parameter with Line Spectrum Pair (LSP). The proposed peaking rate method is measured by the information of LSP in speech. We measured the variety rate of phenomenon about utterances which have different velocity, respectively. As a result, It has distinct variation rate of phenomenon between utterances uttered fast and slow and the rate is 42.8% higher in case of uttered fast than in case of uttered slow.

• MALSORI
• /
• no.44
• /
• pp.105-113
• /
• 2002

In Korean, each digit is monosyllable and some pairs are known to have high confusability, causing performance degradation of connected digit recognition systems. To improve the performance, in this paper, we employ various discriminant analyses (DA) including Linear DA (LDA), Weighted Pairwise Scatter LDA WPS-LDA), Heteroscedastic Discriminant Analysis (HDA), and Maximum Likelihood Linear Transformation (MLLT). We also examine several combinations of various DA for additional performance improvement. Experimental results show that applying any DA mentioned above improves the string accuracy, but the amount of improvement of each DA method varies according to the model complexity or number of mixtures per state. Especially, more than 20% of string error reduction is achieved by applying MLLT after WPS-LDA, compared with the baseline system, when class level of DA is defined as a tied state and 1 mixture per state is used.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.640-643
• /
• 2003

A Walsh function based associative memory is capable of storing m patterns in a single pattern storage space with Walsh encoding of each pattern. Furthermore, each stored pattern can be matched against the stored patterns extremely fast using algorithmic parallel processing. As such, this special type of memory is ideal for real-time processing of large scale information. However this incredible efficiency generates large amount of crosstalk between stored patterns that incurs mis-recognition. This crosstalk is a function of the set of different sequencies [number of zero crossings] of the Walsh function associated with each pattern to be stored. This sequency set is thus optimized in this paper to minimize mis-recognition, as well as to maximize memory saying. In this paper, this Walsh memory has been applied to the problem of face recognition, where PCA is applied to dimensionality reduction. The maximum Walsh spectral component and genetic algorithm (GA) are applied to determine the optimal Walsh function set to be associated with the data to be stored. The experimental results indicate that the proposed methods provide a novel and robust technology to achieve an error-free, real-time, and memory-saving recognition of large scale patterns.

• Journal of Korea Multimedia Society
• /
• v.2 no.3
• /
• pp.297-307
• /
• 1999

A character recognition system, where a large amount of character images arrive continuously in real time, must preprocess character images very quickly. Moreover, information loss due to image trans-formations such as geometric normalization and thinning needs to be minimized especially when character images are small and noisy. Therefore, we suggest a prompt and effective feature extraction method without transforming original images. For this, boundary pixels are defined in terms of the degree in classification, and those boundary pixels are considered selectively in extracting features. The proposed method is tested by a handwritten character recognition and a car plate number recognition. The experiments show that the proposed method is effective in recognition compared to conventional methods. And an overall reduction of execution time is achieved by completing all the required processing by a single image scan.

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.154.2-154
• /
• 2001

This paper presented a adaptation approach based on maximum model distance (MMD) method. This method shares the same framework as they are used for training speech recognizers with abundant training data. The MMD method could adapt to all the models with or without adaptation data. If large amount of adaptation data is available, these methods could gradually approximate the speaker-dependent ones. The approach is evaluated through the phoneme recognition task on the TIMIT corpus. On the speaker adaptation experiments, up to 65.55% phoneme error reduction is achieved. The MMD could reduce phoneme error by 16.91% even when ...