• ETRI Journal
• /
• v.32 no.2
• /
• pp.345-347
• /
• 2010

In this letter, we present a new speech hash function based on the non-negative matrix factorization (NMF) of linear prediction coefficients (LPCs). First, linear prediction analysis is applied to the speech to obtain its LPCs, which represent the frequency shaping attributes of the vocal tract. Then, the NMF is performed on the LPCs to capture the speech's local feature, which is then used for hash vector generation. Experimental results demonstrate the effectiveness of the proposed hash function in terms of discrimination and robustness against various types of content preserving signal processing manipulations.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.72-75
• /
• 2004

In this paper, a finger-click recognition method is proposed to improve the recognition performance for finger-clicking of a wearable input device, called $SCURRY^{TM}$. The proposed method is composed of three parts including feature extraction part, valid click discrimination part, and cross-talk avoidance part. Two types of MEMS inertial sensors are embedded into the wearable input device to measure the angular velocity of a hand (hand movement) and the acceleration rates at the ends of fingers (finger-click motion). The experiment applied to the $SCURRY^{TM}$ device shows the improved stability and performance.

• Journal of Korea Multimedia Society
• /
• v.23 no.9
• /
• pp.1155-1163
• /
• 2020

Deep learning approach is emerging as a new way to improve the accuracy of tree species identification using bark image. However, the approach has not been studied enough because it is confronted with the problem of acquiring a large volume of bark image dataset. This study solved this problem by utilizing a pretrained off-the-shelf DCNN model. It compares the discrimination power of bark features extracted by each DCNN model. Then it extracts the features by using a selected DCNN model and feeds them to a multi-layer perceptron (MLP). We found out that the ResNet50 model is effective in extracting bark features and the MLP could be trained well with the features reduced by the principal component analysis. The proposed approach gives accuracy of 99.1% and 98.4% for BarkTex and Trunk12 datasets respectively.

• Proceedings of the IEEK Conference
• /
• 2003.07c
• /
• pp.2875-2878
• /
• 2003

In this paper, we propose an EEG-based mental state prediction method during a mental tasks. In the experimental task, a subject goes through the process of responding to visual stimulus, understanding the given problem, controlling hand motions, and hitting a key. Considering the subject's varying brain activities, we model subjects' mental states with defining selection time. EEG signals from four subjects were recorded while they performed three mental tasks. Feature vectors defined by these representations were classified with a standard, feed-forward neural network trained via the error back-propagation algorithm. We expect that the proposed detection method can be a basic technology for brain-computer interface by combining with left/right hand movement or cognitive decision discrimination methods.

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10b
• /
• pp.1966-1970
• /
• 1991

The self-mixing type semiconductor laser Doppler velocimeter(SM-LDV) is applied to measure two simultaneously moving targets with different velocities in the same direction as a prototype target for multiscatterers. The measured beat waveform is found to be a composite wave of each beat waveform measured fran each of only moving target. In the composite waveform, each one-cycle wave has a feature of the sawtooth wave. This fact shows a possibility to discriminate the flow direction of fluid containing multiscatterers with distributed velocities by cooperating an improved version of the direction discrimination circuit already devised by the authors.

• Journal of the Korean Institute of Telematics and Electronics S
• /
• v.34S no.2
• /
• pp.94-103
• /
• 1997

The WCHF-FSJTC (wavelet circular harmonic function frequency selective joint transform correlator) using th wavelet transformed CHF as the reference image in FSJTC is proposed for rotation invariant pattern recognition. Since the wavelet transform has the property of feature extraction, the proposed system can have the better DC (discrimination cpability) and the higher SNR(signal to noise ratio) compared with the conventional CHF-CJTC(circular harmonic function conventional joint transform correlator). And since the structure of the proposed system is FSJTC which can eliminate auto-correlation and cross-correlation between input images, it can eliminate false alarm caused by the overlapping among correlation peaks. The used wavelet functio is the morlet function, which is proper for the reference image used in this paper. the optimal dialation parameter and oscillation frequency of the wavelet function are also achieved with varying the parameters of the wavelet function. The computer simulation shows that the proposed system has the best performance when the dilation parameter is 0.8 and the oscillation frequency is 0.48.

• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.283-285
• /
• 2005

To account for the importance of variable in multi-sensor data fusion, random forests are applied to supervised land-cover classification. The random forests approach is a non-parametric ensemble classifier based on CART-like trees. Its distinguished feature is that the importance of variable can be estimated by randomly permuting the variable of interest in all the out-of-bag samples for each classifier. Supervised classification with a multi-sensor remote sensing data set including optical and polarimetric SAR data was carried out to illustrate the applicability of random forests. From the experimental result, the random forests approach could extract important variables or bands for land-cover discrimination and showed good performance, as compared with other non-parametric data fusion algorithms.

• Journal of Broadcast Engineering
• /
• v.7 no.3
• /
• pp.262-270
• /
• 2002

Although most image indexing schemes ate based on global image features, they have limited discrimination capability because they cannot capture local variations of the image. In this paper, we propose a new region-based image retrieval system that can extract important regions in the image using salient point extraction and image segmentation techniques. Our experimental results show that color and texture information in the region provide a significantly improved retrieval performances compared to the global feature extraction methods.

• Journal of Korea Multimedia Society
• /
• v.21 no.8
• /
• pp.917-924
• /
• 2018

Recently, researches using LBP and SVM have been performed as one of the image - based methods for facial emotion recognition. LBP, introduced by Ojala et al., is widely used in the field of image recognition due to its high discrimination of objects, robustness to illumination change, and simple operation. In addition, CS(Center-Symmetric)-LBP was used as a modified form of LBP, which is widely used for face recognition. In this paper, we propose a method to detect four facial expressions such as expressionless, happiness, surprise, and anger using deep neural network. The validity of the proposed method is verified using accuracy. Based on the existing LBP feature parameters, it was confirmed that the method using the deep neural network is superior to the method using the Adaboost and SVM classifier.

• Proceedings of the KIEE Conference
• /
• 2000.11d
• /
• pp.761-764
• /
• 2000

This paper presents a Fuzzy Neural Network (FNN) system to decide whether or not the right information of lanes can be extracted from gray-level images of road scene. The decision of noisy level of input images has been required because much noises usually deteriorates the performance of feature detection based on image processing and lead to erroneous results. As input parameters to FNN, eight noisiness indexes are constructed from a cumulative distribution function (CDF) and proved the indexes being classifiers of images as the good and the bad corrupted by sources of noise by correlation analysis between input images and the indexes. Considering real-time processing and discrimination efficiency, the proposed FNN is structured by eight input parameters, three fuzzy variables and single output. We conduct much experiments and show that our system has comparable performance in terms of false-positive rates.