• Title/Summary/Keyword: Source separation

Search Result 450, Processing Time 0.023 seconds

Frequency Range Enhancement for Faster Convergence of Neural Music Source Separation Systems (신경망 기반 음원 분리 시스템의 학습 속도 향상을 위한 음역대 강조 기법)

  • Kim, Min-Seok;Choi, Woo-Sung;Jung, Soon-Young
    • Proceedings of the Korea Information Processing Society Conference
    • /
    • 2020.05a
    • /
    • pp.567-569
    • /
    • 2020
  • 여러 악기가 섞여 있는 음원으로부터 원하는 악기 소리를 추출하는 음원 분리 기법 중 최근 신경망 기반 시스템이 활발히 연구되고 있다. 악기마다 고유의 음역대를 가진다는 사실에 감안하여, 연구진은 기존 음원 분리 신경망에 적은 수의 학습 파라미터를 추가하여 학습 속도를 대폭 향상시킬 수 있는 음역대 강조 기법을 제안한다.

A sturdy on the blind audio source separation based on multi-step NMF-EM algorithm (다중 단계 NMF-EM 알고리즘 기반의 오디오 소스 분리 방법에 대한 연구)

  • Cho, Choongsang;Kim, Jewoo
    • Proceedings of the Korean Society of Broadcast Engineers Conference
    • /
    • 2014.06a
    • /
    • pp.9-11
    • /
    • 2014
  • 본 논문에서는 오디오 신호의 특성 표현에 유용한 nonnegative matrix factorization(NMF)에 대해 설명하였으며, expectation maximization (EM)을 이용한 NMF 파라미터 추출 및 EM-NMF 기반한 오디오 소스 분리 기술에 대해서 설명했다. 또한, 다중 단계 NMF-EM 구조의 객체 분리를 통해서 객체 분리 성능을 향상시키기 위한 알고리즘을 제안하며, 제안된 알고리즘은 K-pop 음원과 SDR(source distortion ratio)를 통해서 객체 분리 성능을 평가한다. 성능 평가 결과 제안된 알고리즘은 다중 단계를 통해 약 3dB 의 보컬 분리 성능이 향상되며, 상업적 음원 제작에서 사용되는 가상 오디오 효과가 많이 적용된 음원에서 약 5dB 의 분리 성능을 향상시켰다. 그러므로 제안된 방식은 오디오 객체 분리에 유용한 방법이 될 것으로 생각된다.

  • PDF

Modeling of non-isothermal CO2 particle leaked from pressurized source: I. Behavior of single bubble

  • Chang, Daejun;Han, Sang Heon;Yang, Kyung-Won
    • Ocean Systems Engineering
    • /
    • v.2 no.1
    • /
    • pp.17-31
    • /
    • 2012
  • This study investigated the behavior of a non-isothermal $CO_2$ bubble formed through a leak process from a high-pressure source in a deep sea. Isenthalpic interpretation was employed to predict the state of the bubble just after the leak. Three modes of mass loss from the rising bubble were demonstrated: dissolution induced by mass transfer, condensation by heat transfer and phase separation by pressure decrease. A graphical interpretation of the last mode was provided in the pressure-enthalpy diagram. A threshold pressure (17.12 bar) was identified below which the last mode was no longer present. The second mode was as effective as the first for a bubble formed in deep water, leading to faster mass loss. To the contrary, only the first mode was active for a bubble formed in a shallow region. The third mode was insignificant for all cases.

Optimization of an Annular Fin with Variable Pipe Inside Radius in the Heat Exchanger of Ground Source System (지열시스템의 열교환기에서 원 관 내부 반경이 변하는 환형 핀의 최적화)

  • Kang, Hyung-Suk
    • New & Renewable Energy
    • /
    • v.5 no.1
    • /
    • pp.40-46
    • /
    • 2009
  • Optimization of a rectangular profile annular fin with variable pipe inside radius is presented. This optimum procedure is based on fixed fin height and is made by using variables separation method. The optimum heat loss, corresponding optimum fin length and optimum fin efficiency are presented as a function of pipe inside radius, fin half height, inside fluid convection characteristic number and ambient convection characteristic number. One of results shows that the optimum fin length increases linearly with increase of pipe inside radius for fixed fin height and fin base radius.

  • PDF

Experimental and Theoretical Study on Main Noise Sources and Its Radiations of Upwind Wind Turbines (상류형 풍력 터빈의 주요 소음원과 방사소음에 대한 실험적/이론적 고찰)

  • Lee, Gwang-Se;Cheong, Cheol-Ung;Shin, Su-Hyun;Jung, Sung-Soo
    • Proceedings of the Korean Society for Noise and Vibration Engineering Conference
    • /
    • 2010.05a
    • /
    • pp.72-73
    • /
    • 2010
  • In this paper, the main noise sources and its radiated noise levels of upwind horizontal-axis wind turbines are experimentally and theoretically investigated. Theoretical predictions for indentifying the dominant source locations are made by using the empirical noise prediction model of Brooks et al. (1989) for the airfoil self noise. Through the comparison of theoretical results with the experimental results, turbulence-boundary-layer-trailing-edge (TBL-TE) noise is revealed to be the dominant source over all frequency range and separation and stall (S-S) noise is possibly important in the relative lower frequency range compared with TBL-TE noise.

  • PDF

Transport of Metal Ions Across Bulk Liquid Membrane by Lipophilic Acyclic Polyether Dicarboxylic Acids (Lipophilic Acyclic Polyether Dicarboxylic Acid 에 의한 액체막을 통한 금속이온의 이동)

  • Jo, Mun Hwan;Jo, Seong Ho;Lee, In Jong
    • Journal of the Korean Chemical Society
    • /
    • v.38 no.2
    • /
    • pp.129-135
    • /
    • 1994
  • Acyclic polyether dicarboxylic acid have been studied as metal cation carriers in a bulk liquid membrane system. The proton-ionizable ligands feature allows the coupling of a cation transport to reverse proton transport. This feature offers promise for the effective separation and concentration of metal cations with the metal cation transport being driven by a pH gradient. Metal cation transport increased regularly with increasing hydroxide($H^-$) concentration of source phase and with proton($H^+$) concentration of receiving phase. Competitive transport by the acyclic polyether dicarboxylic acids is selective for calcium ion over other alkaline-earth cations.

  • PDF

Performance Analysis of a Combined Cycle of Kalina and Absorption Refrigeration for Recovery of Low-Temperature Heat Source (저온 열원의 활용을 위한 칼리나/흡수냉동 복합사이클의 성능 해석)

  • KIM, KYOUNG HOON;KO, HYUNG JONG;JUNG, YOUNG GUAN
    • Transactions of the Korean hydrogen and new energy society
    • /
    • v.29 no.5
    • /
    • pp.490-496
    • /
    • 2018
  • Recently, the power and refrigeration cogeneration based on Kalina cycle has attracted much attention for more efficient utilization of low-grade energy. This study presents a thermodynamic performance analysis of a cogeneration cycle of power and absorption refrigeration based on Kalina cycle. The cycle combines Kalina cycle (KCS-11) and absorption cycles by adding a condenser and an evaporator between turbine and absorber. The effects of ammonia mass fraction and separation pressure were investigated on the system performance of the system. Results showed that the energy utilization of the system could be greatly improved compared to the basic Kalina cycle.

Study on 3-D Physical Modeling for the Inspection of Tunnel Lining Structure by using Ultrasonic Reflection Method (터널 지보구조 진단을 위한 초음파 반사법을 이용한 3차원 모형실험 연구)

  • 김중열;김유성;신용석;현혜자
    • Proceedings of the Korean Geotechical Society Conference
    • /
    • 2002.03a
    • /
    • pp.221-228
    • /
    • 2002
  • Thickness of concrete lining, voids at the back of lining or shotcrete are very important elements for inspecting the safety of tunnels. Therefore, the inspection of tunnel lining structure means to investigate the inner layer boundaries of the structure. For this purpose, seismic reflection survey is the most desirable method if it works in good conditions. However, the conventional seismic reflection method can not be properly used for investigating thin layers in the lining structure. In other words, to detect the inner boundaries, it is desirable for the wavelength of source to be less than the thickness of each layer and for the receiver to be capable of detecting high frequency(ultrasonic) signals. To this end, new appropriate source and receiver devices should be developed above all for the ultrasonic reflection survey. This paper deals primarily with the development of source and receiver devices which are essential parts of field measuring system. Interests are above all centered in both the radiation pattern, energy, frequency content of the source and the directional sensitivity of the receiver. With these newly devised ones, ultrasonic physical modeling has been performed on 3-D physical model composed of bakelite, water-proof and concrete, The measured seismograms showed a clear separation of wave arrivals reflected from each layer boundary. Furthermore, it is noteworthy that reflection events from the bottom of concrete below water-proof could be also observed. This result demonstrates the usefulness of the both devices that can be applied to benefit the ultrasonic reflection survey. Future research is being focus on dealing with at first an optimal configuration of source and receiver devices well coupled to tunnel wall, and further an efficient data control system of practical use.

  • PDF

Two-Microphone Binary Mask Speech Enhancement in Diffuse and Directional Noise Fields

  • Abdipour, Roohollah;Akbari, Ahmad;Rahmani, Mohsen
    • ETRI Journal
    • /
    • v.36 no.5
    • /
    • pp.772-782
    • /
    • 2014
  • Two-microphone binary mask speech enhancement (2mBMSE) has been of particular interest in recent literature and has shown promising results. Current 2mBMSE systems rely on spatial cues of speech and noise sources. Although these cues are helpful for directional noise sources, they lose their efficiency in diffuse noise fields. We propose a new system that is effective in both directional and diffuse noise conditions. The system exploits two features. The first determines whether a given time-frequency (T-F) unit of the input spectrum is dominated by a diffuse or directional source. A diffuse signal is certainly a noise signal, but a directional signal could correspond to a noise or speech source. The second feature discriminates between T-F units dominated by speech or directional noise signals. Speech enhancement is performed using a binary mask, calculated based on the proposed features. In both directional and diffuse noise fields, the proposed system segregates speech T-F units with hit rates above 85%. It outperforms previous solutions in terms of signal-to-noise ratio and perceptual evaluation of speech quality improvement, especially in diffuse noise conditions.

Determination of Source Contribution Based on Aerosol Number and Mass Concentration in the Seoul Subway Stations (분진의 개수농도 및 질량농도에 입각한 서울시 지하철 역사 내 오염원의 기여도 결정)

  • 최형욱;황인조;김신도;김동술
    • Journal of Korean Society for Atmospheric Environment
    • /
    • v.20 no.1
    • /
    • pp.17-31
    • /
    • 2004
  • The subway play an important part in serious traffic problems. However, because subway system is a closed environment, many serious air pollution problems occurred in subway stations and injured passenger's health. Therefor, it is a necessary to identify sources and to estimate pollutant sources in order to protect passenger's health and to keep clean subway environment. The purpose of this study was to analyze a air quality in the subway stations and to apply a new receptor methodology for quantitatively estimate of PM10 sources. In this study, the size distributions of particulate matters has been measured by using Aerosizer LD (U.S.A., API, Inc.). It's real time measurement capability of time-of-flight technique offers a significant advantage of user convenience and air pollution management. Also, the mass concentrations of PM 10 has been measured by using mini-vol portable sampler (U.S.A., Airmetrics Co.). The sampling performed in Seoul subway stations during the period of February 2000 and April 2000. The number distribution data used in this study consisted of 26 raw data sets in the Jongno-sam-ga station. Correlation Analysis can be used in subway stations for source separation and identification. Then, number contribution from each source is determined by the particle number balance (PNB). The mass concentration data used in this study consisted of 31 raw data in the 8 different stations. The mass contributions of PM10 sources in the concourse by using PMF/CMB model.