• ETRI Journal
• /
• v.43 no.1
• /
• pp.82-94
• /
• 2021

In this work, six voiced/unvoiced speech classifiers based on the autocorrelation function (ACF), average magnitude difference function (AMDF), cepstrum, weighted ACF (WACF), zero crossing rate and energy of the signal (ZCR-E), and neural networks (NNs) have been simulated and implemented in real time using the TMS320C6713 DSP starter kit. These speech classifiers have been integrated into a linear-predictive-coding-based speech analysis-synthesis system and their performance has been compared in terms of the percentage of the voiced/unvoiced classification accuracy, speech quality, and computation time. The results of the percentage of the voiced/unvoiced classification accuracy and speech quality show that the NN-based speech classifier performs better than the ACF-, AMDF-, cepstrum-, WACF- and ZCR-E-based speech classifiers for both clean and noisy environments. The computation time results show that the AMDF-based speech classifier is computationally simple, and thus its computation time is less than that of other speech classifiers, while that of the NN-based speech classifier is greater compared with other classifiers.

• Journal of Information Processing Systems
• /
• v.9 no.1
• /
• pp.117-140
• /
• 2013

Hindi is the most widely spoken language in India, with more than 300 million speakers. As there is no separation between the characters of texts written in Hindi as there is in English, the Optical Character Recognition (OCR) systems developed for the Hindi language carry a very poor recognition rate. In this paper we propose an OCR for printed Hindi text in Devanagari script, using Artificial Neural Network (ANN), which improves its efficiency. One of the major reasons for the poor recognition rate is error in character segmentation. The presence of touching characters in the scanned documents further complicates the segmentation process, creating a major problem when designing an effective character segmentation technique. Preprocessing, character segmentation, feature extraction, and finally, classification and recognition are the major steps which are followed by a general OCR. The preprocessing tasks considered in the paper are conversion of gray scaled images to binary images, image rectification, and segmentation of the document's textual contents into paragraphs, lines, words, and then at the level of basic symbols. The basic symbols, obtained as the fundamental unit from the segmentation process, are recognized by the neural classifier. In this work, three feature extraction techniques-: histogram of projection based on mean distance, histogram of projection based on pixel value, and vertical zero crossing, have been used to improve the rate of recognition. These feature extraction techniques are powerful enough to extract features of even distorted characters/symbols. For development of the neural classifier, a back-propagation neural network with two hidden layers is used. The classifier is trained and tested for printed Hindi texts. A performance of approximately 90% correct recognition rate is achieved.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.6 no.1
• /
• pp.76-82
• /
• 2002

In this paper, we implemented on the classifier for differential diagnosis of laryngeals disease which is normal, polyp, nodule, palsy, and each step of glottic cancer using hierarchical neural network. We conducted on classifier of various vowels as /a/, /e/, /i/, /o/, /u/ from normal group, laryngeal disease group, each step of cancer group. The experimental result on classification of each vowels as follows. A /a/ vowel shows excellent classification result to the other vowels in regard to each Input parameters. Thus we implemented the hierarchical neural network for differential diagnosis of laryngeals disease using only /a/ vowel. A implemented hierarchical neural network is composed of each other laryngeals disease apply to each other parameter in each hierarchical layer. We take the voice signals from patient who get the laryngeal disease and glottic cancer, and then use the APQ, PPQ, vAm, Jitter, Shimmer, RAP as input parameter of neural networks.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.2
• /
• pp.785-804
• /
• 2017

Due to various factors such as postures, facial expressions and illuminations, face recognition by videos often suffer from poor recognition accuracy and generalization ability, since the within-class scatter might even be higher than the between-class one. Herein we address this problem by proposing a hierarchical cascaded classifier for video face recognition, which is a multi-layer algorithm and accounts for the misclassified samples plus their similar samples. Specifically, it can be decomposed into single classifier construction and multi-layer classifier design stages. In single classifier construction stage, classifier is created by clustering and the number of classes is computed by analyzing distance tree. In multi-layer classifier design stage, the next layer is created for the misclassified samples and similar ones, then cascaded to a hierarchical classifier. The experiments on the database collected by ourselves show that the recognition accuracy of the proposed classifier outperforms the compared recognition algorithms, such as neural network and sparse representation.

• Journal of the Institute of Convergence Signal Processing
• /
• v.2 no.2
• /
• pp.1-6
• /
• 2001

Eigenface method in face recognition is useful due to its insensitivity to large variations in facial expression and facial details. However its low recognition rate necessitates additional researches. In this paper, we present an efficient method for improving the recognition rate in face recognition using eigenface feature. For this, we performs a comparative study of three different classifiers which are i) a single prototype (SP) classifier, ii) a nearest neighbor (NN) classifier, and iii) a standard feedforward neural network (FNN) classifier. By evaluating and analyzing the performance of these three classifiers, we shows that the distribution of eigenface features of face image is not compact and that selections of classifier and sample training data are important for obtaining higher recognition rate. Our experiments with the ORL face database show that 1-NN classifier outperforms the SP and FNN classifiers. We have achieved a recognition rate of 91.0% by selecting sample trainging data properly and using 1-NN classifier.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.9 no.5
• /
• pp.517-525
• /
• 1999

In this paper, we propose efficient total recognition system of handwritten and printed numerals for reducing the classification time. The proposed system consists of two-step neuroclassifier : Printed numerals classifier and handwritten numerals classifier. In the proposed scheme, the printed numerals classifier classifies the printed numerals rapidly with single MLP neural network by low-order feature vector and rejects handwritten numerals. The handwritten numerals classifier classifies the handwritten numerals which is rejected in printed numerals classifier with modularized cluster neural network by complex feature vector. In order to verify the performance of the proposed method,handwritten numerals database of NIST and printed numerals database which include various fonts are used in the experiments. In case of using the proposed classifier, the overall classification time was reduced by 49.1% - 65.5% in comparison of the existent handwritten classifier.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2013.05a
• /
• pp.321-322
• /
• 2013

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.2226-2229
• /
• 2003

A new method is presented to construct a classifier. This was accomplished by combining a generalized regression neural network (GRNN) and a genetic algorithm (GA). The classifier constructed in this way is referred to as a GA-GRNN. The GA played a role of controlling training factors simultaneously. In GA optimization, neuron spreads were represented in a chromosome. The proposed optimization method was applied to a data set, consisted of 4 different promoter sequences. The training and test data were composed of 115 and 58 sequence patterns, respectively. The range of neuron spreads was experimentally varied from 0.4 to 1.4 with an increment of 0.1. The GA-GRNN was compared to a conventional GRNN. The classifier performance was investigated in terms of the classification sensitivity and prediction accuracy. The GA-GRNN significantly improved the total classification sensitivity compared to the conventional GRNN. Also, the GA-GRNN demonstrated an improvement of about 10.1% in the total prediction accuracy. As a result, the proposed GA-GRNN illustrated improved classification sensitivity and prediction accuracy over the conventional GRNN.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.27 no.8
• /
• pp.1239-1249
• /
• 1990

In this paper, using the neural network design techniques, an adaptive Mahalanobis distance classifier(AMDC) is designed. This classifier has three layers: input layer, internal layer and output layer. The connection from input layer to internal layer is fully connected, and that from internal to output layer has partial connection that might be thought as an Oring. If two ormore clusters of patterns of one class are laid apart in the feature space, the network adaptively generate the internal nodes, whhch are corresponding to the subclusters of that class. The number of the output nodes in just same as the number of the classes to classify, on the other hand, the number of the internal nodes is defined by the number of the subclusters, and can be optimized by itself. Using the method of making the subclasses, the different patterns that are of the same class can easily be distinguished from other classes. If additional training is needed after the completion of the traning, the AMDC does not have to repeat the trainging that has already done. To test the performance of the AMDC, the experiments of classifying 500 Hangeuls were done. In experiment, 20 print font sets of Hangeul characters(10,000 cahracters) were used for training, and with 3 sets(1,500 characters), the AMDC was tested for various initial variance \ulcornerand threshold \ulcorner and compared with other statistical or neural classifiers.

• Journal of IKEEE
• /
• v.22 no.3
• /
• pp.700-705
• /
• 2018

In this paper, we propose the gas classifier based on restricted column energy neural network (RCE-NN) and present its hardware implementation results for real-time learning and classification. Since RCE-NN has a flexible network architecture with real-time learning process, it is suitable for gas classification applications. The proposed gas classifier showed 99.2% classification accuracy for the UCI gas dataset and was implemented with 26,702 logic elements with Intel-Altera cyclone IV FPGA. In addition, it was verified with FPGA test system at an operating frequency of 63MHz.