• The Journal of the Convergence on Culture Technology
• /
• v.6 no.2
• /
• pp.509-514
• /
• 2020

The conventional TTS system consists of several modules, including text preprocessing, parsing analysis, grapheme-to-phoneme conversion, boundary analysis, prosody control, acoustic feature generation by acoustic model, and synthesized speech generation. But TTS system with deep learning is composed of Text2Mel process that generates spectrogram from text, and vocoder that synthesizes speech signals from spectrogram. In this paper, for the optimal Korean TTS system construction we apply Tacotron2 to Tex2Mel process, and as a vocoder we introduce the methods such as WaveNet, WaveRNN, and WaveGlow, and implement them to verify and compare their performance. Experimental results show that WaveNet has the highest MOS and the trained model is hundreds of megabytes in size, but the synthesis time is about 50 times the real time. WaveRNN shows MOS performance similar to that of WaveNet and the model size is several tens of megabytes, but this method also cannot be processed in real time. WaveGlow can handle real-time processing, but the model is several GB in size and MOS is the worst of the three vocoders. From the results of this study, the reference criteria for selecting the appropriate method according to the hardware environment in the field of applying the TTS system are presented in this paper.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.1
• /
• pp.36-45
• /
• 2015

Recently, a lot of studies have been made about the methods used to generate turbulent velocity fields stochastically in order to effectively predict broadband flow noise. Among them, the FRPM (Fast Random Particle Mesh) method which generates turbulence with specific statistical properties using turbulence kinetic energy and dissipation obtained from the steady solution of the RANS (Reynolds Averaged Navier-Stokes) equations has been successfully applied. However, the FRPM method cannot be applied to the flow noise problems involving intrinsic unsteady characteristics such as centrifugal fan. In this paper, to effectively predict the broadband noise generated by centrifugal fan, U-FRPM (unsteady FRPM) method is developed by extending the FRPM method to be combined with the unsteady numerical solutions of the unsteady RANS equations to generate the turbulence considered as broadband noise sources. Firstly, an unsteady flow field is obtained from the unsteady RANS equations through CFD (Computational Fluid Dynamics). Then, noise sources are generated using the U-FRPM method combined with acoustic analogy. Finally, the linear propagation model which is realized through BEM (Boundary Element Method) is combined with the generated sources to predict broadband noise at the listeners' position. The proposed technique is validated to compare its prediction result with the measured data.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.21 no.6
• /
• pp.751-758
• /
• 2015

Noise prediction for HVAC(Heating, Ventilating and Air Conditioning) systems are normally performed by empirical method suggested by NEBB(National Environmental Balancing Bureau, 1994). However, the method is not suitable for large ducts in ships. In this paper, computational analysis methods are used to develop a noise prediction method for the large ducts in ships. To develop regression formula of attenuation of sound pressure level in large ducts, Boundary Element Method(BEM) is used. BEM and Computational Fluid Dynamics(CFD) are applied to the analysis of flow-induced noise in ducts with stiffeners inside. Loud noise above 100 dB can be generated in some cases. Breakout noises of large ducts are also analyzed by using BEM and Finite Element Method(FEM). The acoustic pressure level shows about 10-15dB difference between inside and outside of the duct. Utilizing the results of this study, it is expected that shipyard planners can predict noise of the HVAC system for ships.

• The Journal of the Acoustical Society of Korea
• /
• v.31 no.7
• /
• pp.461-471
• /
• 2012

The purpose of this paper is to provide basic data for development of Korean teaching programs for immigrant women from multicultural families through the acoustic analysis of their prosodic phrasing and intonation pattern. The results showed that immigrant women showed some differences in most of the prosodic characteristics from a Korean women's group: Immigrant women realized the first word of a sentence in an intonational phrase while Korean women did in an accentual phrase. They also haven't yet correctly learned the tone type of the first of an accentual phrase which differs depending on the type of its first segment yet. As a result, they showed many diverse intonation patterns compared to Korean women. Furthermore, the immigrant women's groups showed some differences between them in a few prosodic characteristics. Philippine women, whose residence duration in Korea is relatively longer than that of Vietnamese women, were more similar to Korean women: Vietnamese women read a sentence with a larger number of intonational phrases than Philippine women did. And they realized sentence-final boundary tone of a yes-no question not only in 'H%' but also in 'HL%' while, like Korean women, Philippine women did only in 'H%'.

• Journal of the Korean Society for Nondestructive Testing
• /
• v.29 no.4
• /
• pp.308-315
• /
• 2009

Because the attenuation of AE signal in composite materials is relatively higher than that of metallic materials, it is required to develop a damage assessment technique less affected by the attenuation property of composite materials in order to use AE sensing as a damage detection method. In the signal processing procedure, it is profitable to use the leading wave that arrives first because the leading wave is less influenced by the boundary conditions. Using wavelet transform, we investigated the frequency characteristics of impact induced AE signals focused on the leading wave in advance and chose the key factors to discriminate the damaged condition quantitatively. In this research, we established a damage assessment technique using the sharing percentage of the wavelet detail components of AE signal, and conducted a low-velocity impact test on composite laminates to confirm the feasibility of the proposed signal processing method.

• The Journal of the Acoustical Society of Korea
• /
• v.34 no.3
• /
• pp.219-226
• /
• 2015

In this paper, derivation of the STL (Sound Transmission Loss) of a square plate installed in an impedance tube is discussed using an analytic method. Coupled motion of the plate vibration and acoustic field is considered. Vibration of the plate and pressure field inside the tube are expressed in terms of the infinite series of modal functions. Under the plane wave assumption, it is shown that consideration of the first few modes yields sufficiently accurate results. When the boundary of the plate is clamped, vibration mode is assumed as a multiplication of the beam modes corresponding to the crosswise directions. The natural frequencies of the clamped plate are calculated using the Rayleigh-Ritz method. It is found that the STL shows a dip at the lowest natural frequency of the plate, and increases as the frequency decreases below the natural frequency. Comparison of the result in this paper with the STL obtained by measurements and FE computations in the reference shows an excellent agreement.

• Proceedings of the KSRS Conference
• /
• 2008.10a
• /
• pp.340-343
• /
• 2008

The Soya Warm Current (SWC) is a coastal boundary current, which flows along the coast of Hokkaido in the Sea of Okhotsk. Seasonal and subinertial variations in the SWC are investigated using data obtained by high-frequency (HF) ocean radars, coastal tide gauges, and a bottom-mounted acoustic Doppler current profiler (ADCP). The HF radars clearly capture the seasonal variations in the surface current fields of the SWC. The velocity of the SWC reaches its maximum, approximately 1 m/s, in the summer, and becomes weaker in the winter. The velocity core is located 20 to 30 km from the coast, and its width is approximately 50 km. The almost same seasonal cycle was repeated in the period from August 2003 to March 2007. In addition to the annual variation, the SWC exhibits subinertial variations with a period from 10-15 days. The surface transport by the SWC shows a significant correlation with the sea level difference between the Sea of Japan and Sea of Okhotsk for both of the seasonal and subinertial variations, indicating that the SWC is driven by the sea level difference between the two seas. Generation mechanism of the subinertial variation is discussed using wind data from the European Centre for Medium-range Weather Forecasts (ECMWF) analyses. The subinertial variations in the SWC are significantly correlated with the meridional wind component over the region. The subinertial variations in the sea level difference and surface current delay from the meridional wind variations for one or two days. Continental shelf waves triggered by the meridional wind on the east coast of Sakhalin and west coast of Hokkaido are considered to be a possible generation mechanism for the subinertial variations in the SWC.

• 대한공업교육학회지
• /
• v.34 no.1
• /
• pp.238-251
• /
• 2009

The noise itself that affects the sensor array is defined as the noise which happens in the place where the system is installed and the circumference noise which comes from the ocean. The array structure for detecting acoustic signal in the underwater effected turbulent layer flow noise. In this paper to design the conformal array spectral density function was introduced and several cases of flow induced noise which affect transfer function were simulated. Modified Corcos wall pressure model was used as turbulent boundary layer flow noise. The effect of noise has been reduced as integrated sum of transfer function has been reduced by decreasing elastomer thickness and density when kx is in low wave number area. Also the characteristics of transfer function by Corcos wall pressure displayed the product of frequency density function. This simulation results can be applied to the conformal array design in unmmaned underwater vehicle in the near future.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.25 no.4
• /
• pp.540-545
• /
• 2001

Many previous researches on the premixed flame in a tube have treated the unsteady flame behaviors in which the shape, position and intensity of the flame varied, but more detail and fundamental research has been necessary. The flame stabilization condition in a tube, a unique steady state, and the unsteady behaviors, using the stabilization condition as an initial condition, were carried out in recent years. In this paper, propane-air premixed flame was stabilized in a tube and the flame behavior was observed when the mean velocity variation was imposed into the opposite direction of the initial mean velocity. The velocity variation is larger than the burning velocity and longer than the reaction time scale. During the period of the velocity variation flame is not extinguished. But after the period of the mean velocity variation the flame could be re-stabilized or be extinguished depending on the experimental conditions: equivalence ratio, period of velocity variation and magnitude of velocity variation. The extinction mechanisms were classified into the two cases, one is caused by the flame stretch in the shear layer near the wall, and the other is caused by the vortices and vortexes, which are generted by the acoustic waves.

• Ocean and Polar Research
• /
• v.23 no.2
• /
• pp.141-160
• /
• 2001

The challenge of predicting the Japanese coastal ocean motivated Frontier Observational Research System for Global Change (FORSGC) and the Japan Marine Science and Technology Center (JAMSTEC) to start a multiyear observational programme in the upstream Kuroshio in November 2000. This field effort, the Kuroshio Observation Program (KOP), should enable us to determine the barotropic and baroclinic components of the western boundary current system, thus, to better understand interactions of the currents with mesoscale eddies, the Kuroshio instabilities, and path bimodality. We, then, will be able to improve modeling predictability of the mesoscale, seasonal, and inter-annual processes in the midstream Kuroshio near the Japanese main islands by using this knowledge. The KOP is focused on an enhanced regional coverage of the sea surface height variability and the baroclinic structure of the mainstream Kuroshio in the East China Sea, the Ryukyu Current east of the Ryukyu's, and the Kuroshio recirculation. An attractive approach of the KOP is a development of a new data acquisition system via acoustic telemetry of the observational data. The monitoring system will provide observations for assimilation into extensive numerical models of the ocean circulation, targeting the real-time monitoring of the Japanese coastal waters.