• Phonetics and Speech Sciences
• /
• v.4 no.2
• /
• pp.79-86
• /
• 2012

In this paper, we propose a novel approach to improve histogram equalization for speaker identification. Our method collects all speech features of UBM training data to make a reference distribution. The ranks of the feature vectors are calculated in the sorted list of the collection of the UBM training data and the test data. We use the ranks to perform order-based histogram equalization. The proposed method improves the accuracy of the speaker recognition system with short utterances. We use four kinds of speech databases to evaluate the proposed speaker recognition system and compare the system with cepstral mean normalization (CMN), mean and variance normalization (MVN), and histogram equalization (HEQ). Our system reduced the relative error rate by 33.3% from the baseline system.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.6
• /
• pp.133-140
• /
• 1999

In this study, a monitoring method to detect chip disposal state in drilling system based on neural network was proposed and its performance was evaluated. If chip flow is bad during drilling, not only the static component but also the fluctuation of dynamic component of drilling. Drilling torque is indirectly measured by sensing spindle motor power through a AC spindle motor drive system. Spindle motor power being measured drilling, four quantities such as variance/mean, mean absolute deviation, gradient, event count were calculated as feature vectors and then presented to the neural network to make a decision on chip disposal state. The selected features are sensitive to the change of chip disposal state but comparatively insensitive to the change of drilling condition. The 3 layerd neural network with error back propagation algorithm has been used. Experimental results show that the proposed monitoring system can successfully recognize the chip disposal state over a wide range of drilling condition even though it is trained under a certain drilling condition.

• IEIE Transactions on Smart Processing and Computing
• /
• v.2 no.5
• /
• pp.255-265
• /
• 2013

3DTV is expected to be a promising next-generation broadcasting service. On the other hand, the visual discomfort/fatigue problems caused by viewing 3D videos have become an important issue. This paper proposes a perceptual quality assessment metric for a stereoscopic video (SV-PQAM). To model the SV-PQAM, this paper presents the following features: temporal variance, disparity variation in intra-frames, disparity variation in inter-frames and disparity distribution of frame boundary areas, which affect the human perception of depth and visual discomfort for stereoscopic views. The four features were combined into the SV-PQAM, which then becomes a no-reference stereoscopic video quality perception model, as an objective quality assessment metric. The proposed SV-PQAM does not require a depth map but instead uses the disparity information by a simple estimation. The model parameters were estimated based on linear regression from the mean score opinion values obtained from the subjective perception quality assessments. The experimental results showed that the proposed SV-PQAM exhibits high consistency with subjective perception quality assessment results in terms of the Pearson correlation coefficient value of 0.808, and the prediction performance exhibited good consistency with a zero outlier ratio value.

• The Journal of Korea Robotics Society
• /
• v.16 no.2
• /
• pp.103-111
• /
• 2021

This study addresses a low-cost lidar sensor-based glass feature extraction method for an accurate map representation using statistical moments, i.e. the mean and variance. Since the low-cost lidar sensor produces range-only data without intensity and multi-echo data, there are some difficulties in detecting glass-like objects. In this study, a principle that an incidence angle of a ray emitted from the lidar with respect to a glass surface is close to zero degrees is concerned for glass detection. Besides, all sensor data are preprocessed and clustered, which is represented using statistical moments as glass feature candidates. Glass features are selected among the candidates according to several conditions based on the principle and geometric relation in the global coordinate system. The accumulated glass features are classified according to the distance, which is lastly represented on the map. Several experiments were conducted in glass environments. The results showed that the proposed method accurately extracted and represented glass windows using proper parameters. The parameters were empirically designed and carefully analyzed. In future work, we will implement and perform the conventional SLAM algorithms combined with our glass feature extraction method in glass environments.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.7 no.4
• /
• pp.727-732
• /
• 2012

This paper presents an approach to classify normal and arrhythmia from the MIT-BIH Arrhythmia Database using Discrete Cosine Transform(DCT), Discrete Wavelet Transform(DWT) and neural network. In the first step, Discrete Cosine Transform is used to obtain the representative 15 coefficients for input features of neural network. In the second step, Discrete Wavelet Transform are used to extract maximum value, minimum value, mean value, variance, and standard deviation of detail coefficients. Neural network classifies normal and arrhythmia beats using 55 numbers of input features, and then the accuracy rate is 98.8%.

• The Journal of the Acoustical Society of Korea
• /
• v.31 no.2
• /
• pp.87-92
• /
• 2012

In this paper, a robust feature compensation method to deal with the environmental mismatch is proposed. The proposed method applies energy based weights according to the degree of speech presence to the Mean subtraction, Variance normalization, and ARMA filtering (MVA) processing. The weights are further smoothed by the moving average and maximum filters. The proposed feature compensation algorithm is evaluated on AURORA 2 task and distant talking experiment using the robot platform, and we obtain error rate reduction of 14.4 % and 44.9 % by using the proposed algorithm comparing with MVA processing on AURORA 2 task and distant talking experiment, respectively.

• Communications for Statistical Applications and Methods
• /
• v.23 no.6
• /
• pp.563-574
• /
• 2016

In this study, we propose several simultaneous tests to detect the difference between means and variances for the two-sample problem when the underlying distribution is normal. For this, we apply the likelihood ratio principle and propose a likelihood ratio test. We then consider a union-intersection test after identifying the likelihood statistic, a product of two individual likelihood statistics, to test the individual sub-null hypotheses. By noting that the union-intersection test can be considered a simultaneous test with combination function, also we propose simultaneous tests with combination functions to combine individual tests for each sub-null hypothesis. We apply the permutation principle to obtain the null distributions. We then provide an example to illustrate our proposed procedure and compare the efficiency among the proposed tests through a simulation study. We discuss some interesting features related to the simultaneous test as concluding remarks. Finally we show the expression of the likelihood ratio statistic with a product of two individual likelihood ratio statistics.

• Smart Structures and Systems
• /
• v.17 no.6
• /
• pp.1067-1085
• /
• 2016

Nonstationary features existing in tropical storms have been frequently captured in recent field measurements, and the applicability of the stationary theory to the analysis of wind characteristics needs to be discussed. In this study, a tropical storm called Nakri measured at Taizhou Bridge site based on structural health monitoring (SHM) system in 2014 is analyzed to give a comparison of the stationary and nonstationary characteristics. The stationarity of the wind records in the view of mean and variance is first evaluated with the run test method. Then the wind data are respectively analyzed with the traditional stationary model and the wavelet-based nonstationary model. The obtained wind characteristics such as the mean wind velocity, turbulence intensity, turbulence integral scale and power spectral density (PSD) are compared accordingly. Also, the stationary and nonstationary PSDs are fitted to present the turbulence energy distribution in frequency domain, among which a modulating function is included in the nonstationary PSD to revise the non-monotonicity. The modulated nonstationary PSD can be utilized to unconditionally simulate the turbulence presented by the nonstationary wind model. The results of this study recommend a transition from stationarity to nonstationarity in the analysis of wind characteristics, and further in the accurate prediction of wind-induced vibrations for engineering structures.

• Journal of Applied Reliability
• /
• v.14 no.4
• /
• pp.208-212
• /
• 2014

An experimental gear-box was set-up to simulate the real situation of the wind-turbine. Artificial cracks of different sizes were machined into the gear. Vibration signals were acquired to diagnose the different crack fault conditions. Time-domain features such as root mean square, variance, kurtosis, normalized 6th central moments were used to capture the characteristics of different crack conditions. Normal condition, 1 mm crack condition, 2mm crack condition, 6mm crack condition, and tooth fault condition were compared using ANFIS and DAG-SVM methods, and three different DAG-SVM models were compared. High-pass filtering improved the success rates remarkably in the case of DAG-SVM.

• Journal of Institute of Control, Robotics and Systems
• /
• v.15 no.5
• /
• pp.457-464
• /
• 2009

The weak classifier of AdaBoost algorithm is a central classification element that uses a single criterion separating positive and negative learning candidates. Finding the best criterion to separate two feature distributions influences learning capacity of the algorithm. A common way to classify the distributions is to use the mean value of the features. However, positive and negative distributions of Haar-like feature as an image descriptor are hard to classify by a single threshold. The poor classification ability of the single threshold also increases the number of boosting operations, and finally results in a poor classifier. This paper proposes a weak classifier that uses multiple criterions by adding a probabilistic criterion of the positive candidate distribution with the conventional mean classifier: the positive distribution has low variation and the values are closer to the mean while the negative distribution has large variation and values are widely spread. The difference in the variance for the positive and negative distributions is used as an additional criterion. In the learning procedure, we use a new classifier that provides a better classifier between them by selective switching between the mean and standard deviation. We call this new type of combined classifier the "Mixed Weak Classifier". The proposed weak classifier is more robust than the mean classifier alone and decreases the number of boosting operations to be converged.