• The Journal of the Korea institute of electronic communication sciences
• /
• v.5 no.4
• /
• pp.363-371
• /
• 2010

In an effort to enhance the quality of feature vector classification and thereby reduce the recognition error rate for the speaker-independent speech recognition, we study the effect of spectral tilt on the Fourier magnitude spectrum en route to the extraction of MFCC. The effect of FIR filtering on the speech signal on the speech recognition is also investigated in parallel. Evaluation of the proposed methods are performed by two independent ways of the Fisher discriminant objective function and speech recognition test by hidden Markov model with fuzzy vector quantization. From the experiments, the recognition error rate is found to show about 10% relative improvements over the conventional method by an appropriate choice of the tilt factor.

• Asian-Australasian Journal of Animal Sciences
• /
• v.14 no.6
• /
• pp.858-861
• /
• 2001

Near Infrared spectroscopy was applied to the samples of processed pork to see the effect of grinding on chemical components analyses. Data from conventional chemical analyses of moisture, fat, protein, NaCl were put into calibration model by NIR of reflectance mode. The other properties observed were pH and color parameters ($L^*,\;a^*,\;b^*$). Spectral ranges of 400~2500 nm and 400~1100 nm were compared for color parameters. Spectral ranges of 400~2500 nm and 1100~2500 nm were compared for chemical components and pH. Different spectral ranges caused little changes in the coefficients of determination or standard errors. $R^{2,}s$ of calibration models for color parameters were in the range of 0.97 to 1.00. $R^{2,}s$ of calibration models of intact sausages for moisture, protein, fat, NaCl and pH were 0.98, 0.89, 0.95, 0.73 and 0.77, respectively using spectra at 1100~2500 nm. $R^{2,}s$ of calibration models of ground sausages for moisture, protein, fat, NaCl and pH were 0.97, 0.91, 0.97, 0.42 and 0.56, respectively using spectra at 1100~2500 nm.

• Phonetics and Speech Sciences
• /
• v.9 no.1
• /
• pp.41-47
• /
• 2017

The main advantage of the statistical parametric speech synthesis is its flexibility in changing voice characteristics. A personalized text-to-speech(TTS) system can be implemented by combining a speech synthesis system and a voice transformation system, and it is widely used in many application areas. It is known that the fundamental frequency and the spectral envelope of speech signal can be independently modified to convert the voice characteristics. Also it is important to maintain naturalness of the transformed speech. In this paper, a speech synthesis system based on Hidden Markov Model(HMM-based speech synthesis, HTS) using the STRAIGHT vocoder is constructed and voice transformation is conducted by modifying the fundamental frequency and spectral envelope. The fundamental frequency is transformed in a scaling method, and the spectral envelope is transformed through frequency warping method to control the speaker's vocal tract length. In particular, this study proposes a voice transformation method using the correlation between fundamental frequency and vocal tract length. Subjective evaluations were conducted to assess preference and mean opinion scores(MOS) for naturalness of synthetic speech. Experimental results showed that the proposed voice transformation method achieved higher preference than baseline systems while maintaining the naturalness of the speech quality.

• Proceedings of the Korean Association of Geographic Inforamtion Studies Conference
• /
• 2010.06a
• /
• pp.130-136
• /
• 2010

This study estimates and evaluates the parameters to relate spectral intensities of aerial and terrestrial images through spectral analysis of each band. For the experiment, an aerial image covering the headquater of the Kyungpook National University was used and terrestrial images were taken by the Sony DSC-F828 DSLR camera. For finding the spectral correspondence, gray intensity, RGB variance, mean, standard deviation were computed, from which parameters of a linear model between patches of both images were computed and evaluated using check patches.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.2E
• /
• pp.56-67
• /
• 2004

Despite abundant research outcomes of blind source separation (BSS) in many types of simulated environments, their performances are still not satisfactory to be applied to the real environments. The major obstacle may seem the finite filter length of the assumed mixing model and the nonlinear sensor noises. This paper presents a two-step speech enhancement method with multiple microphone inputs. The first step performs a frequency-domain BSS algorithm to produce multiple outputs without any prior knowledge of the mixed source signals. The second step further removes the remaining cross-channel interference by a spectral cancellation approach using a probabilistic source absence/presence detection technique. The desired primary source is detected every frame of the signal, and the secondary source is estimated in the power spectral domain using the other BSS output as a reference interfering source. Then the estimated secondary source is subtracted to reduce the cross-channel interference. Our experimental results show good separation enhancement performances on the real recordings of speech and music signals compared to the conventional BSS methods.

• Structural Engineering and Mechanics
• /
• v.28 no.4
• /
• pp.387-410
• /
• 2008

The fatigue damage accumulation rates of horizontally curved thin walled box-girder bridge have been estimated from vehicle-induced dynamic stress history using rain flow cycle counting method in the time domain approach. The curved box-girder bridge has been numerically modeled using computationally efficient thin walled box-beam finite elements, which take into account the important structural actions like torsional warping, distortion and distortional warping in addition to the conventional displacement and rotational degrees of freedom. Vehicle model includes heave-pitch-roll degrees of freedom with longitudinal and transverse input to the wheels. The bridge deck unevenness, which is taken as inputs to the vehicle wheels, has been assumed to be a realization of homogeneous random process specified by a power spectral density (PSD) function. The linear damage accumulation theory has been applied to calculate fatigue life. The fatigue life estimated by cycle counting method in time domain has been compared with those found by estimating the PSD of response in frequency domain. The frequency domain method uses an analytical expression involving spectral moment characteristics of stress process. The effects of some of the important parameters on fatigue life of the curved box bridge have been studied.

• Wind and Structures
• /
• v.29 no.6
• /
• pp.389-403
• /
• 2019

In this study, two original spectral representations of stationary stochastic fields, say the continuous proper orthogonal decomposition (CPOD) and the frequency-wavenumber spectral representation (FWSR), are derived from the Fourier-Stieltjes integral at first. Meanwhile, the relations between the above two representations are discussed detailedly. However, the most widely used conventional Monte Carlo schemes associated with the two representations still leave two difficulties unsolved, say the high dimension of random variables and the incompleteness of probability with respect to the generated sample functions of the stochastic fields. In view of this, a dimension-reduction model involving merely one elementary random variable with the representative points set owing assigned probabilities is proposed, realizing the refined description of probability characteristics for the stochastic fields by generating just several hundred representative samples with assigned probabilities. In addition, for the purpose of overcoming the defects of simulation efficiency and accuracy in the FWSR, an improved scheme of non-uniform wavenumber intervals is suggested. Finally, the Fast Fourier Transform (FFT) algorithm is adopted to further enhance the simulation efficiency of the horizontal stochastic wind velocity fields. Numerical examplesfully reveal the validity and superiorityof the proposed methods.

• Journal of Information Processing Systems
• /
• v.16 no.2
• /
• pp.435-446
• /
• 2020

With the ability to learn rules from training data, the machine learning model can classify unknown objects. At the same time, the dimension of hyperspectral data is usually large, which may cause an over-fitting problem. In this research, an identification methodology of tea diseases was proposed based on spectral reflectance and machine learning, including the feature selector based on the decision tree and the tea disease recognizer based on random forest. The proposed identification methodology was evaluated through experiments. The experimental results showed that the recall rate and the F1 score were significantly improved by the proposed methodology in the identification accuracy of tea disease, with average values of 15%, 7%, and 11%, respectively. Therefore, the proposed identification methodology could make relatively better feature selection and learn from high dimensional data so as to achieve the non-destructive and efficient identification of different tea diseases. This research provides a new idea for the feature selection of high dimensional data and the non-destructive identification of crop diseases.

• Bulletin of the Korean Chemical Society
• /
• v.26 no.1
• /
• pp.115-118
• /
• 2005

The miniaturized NIR (Near-infrared) spectrometer has been utilized for the determination of water content (1-19% range) in ethanol that is the most popular organic solvent in pharmaceutical industries. It has many potential capabilities that can replace the conventional analyzers especially for the on-line measurement since it is compact, versatile and cost-effective. By using two dimensional (2D) correlation spectroscopy, it was preliminarily investigated to find any unforeseen spectral distortion among the spectra collected from the miniaturized spectrometer. The 2D study revealed that the spectral variation clearly followed the variation of water concentration without any spectral distortion or abnormality. PLS (Partial Least Squares) was employed to build the calibration model and the resulting prediction performance was acceptable and stable over several days. Even though the miniaturized NIR system was evaluated to fairly simple chemical matrix, the overall study demonstrates the sufficient feasibility for diverse practical and industrial applications.

• Journal of Biosystems Engineering
• /
• v.36 no.3
• /
• pp.180-186
• /
• 2011

The objective of this research was to establish the estimation method of growth information on potato using Multi-Spectral Image Sensor (MSIS) and Global Positioning System (GPS). And growth estimation map for determining a prescription map over the entire field was generated. To determine the growth model, 10 ground-truth points of areas of $4m^2$ each were selected and investigated. The growth information included stem number, crop height and SPAD value. In addition, images information involving the ground-truth points were also taken by an unmanned helicopter, and reflectance value of Green, Red, and NIR bands were calculated with image processing. Then, growth status of potato was modeled by multi-regression analysis using these reflectance value of Green, Red, and NIR. As a result, potato growth information could be detected by analyzing Green, Red, and NIR images. Stem number, crop height and SPAD value could be estimated with $R^2$ values of 0.600, 0.657 and 0.747 respectively. The generated GIS map would describe variability of the potato growth in a whole field.