• ETRI Journal
• /
• 제39권1호
• /
• pp.21-29
• /
• 2017

This paper addresses the problem of unsupervised speech separation based on robust non-negative matrix factorization (RNMF) with ${\beta}$-divergence, when neither speech nor noise training data is available beforehand. We propose a robust version of non-negative matrix factorization, inspired by the recently developed sparse and low-rank decomposition, in which the data matrix is decomposed into the sum of a low-rank matrix and a sparse matrix. Efficient multiplicative update rules to minimize the ${\beta}$-divergence-based cost function are derived. A convolutional extension of the proposed algorithm is also proposed, which considers the time dependency of the non-negative noise bases. Experimental speech separation results show that the proposed convolutional RNMF successfully separates the repeating time-varying spectral structures from the magnitude spectrum of the mixture, and does so without any prior training.

• Journal of Mechanical Science and Technology
• /
• 제17권2호
• /
• pp.280-287
• /
• 2003

The unsteady flow characteristics and the related noise of separated incompressible laminar boundary layer flows (Re$\sub$$\delta$/* = 614, 868, and 1,063) are numerically investigated. The characteristic lines of the wall pressure are examined to identify the primary noise source, related with the unsteady motion of the vortex at the reattachment point of the separation bubble. The generation and propagation of the vortex-induced noise in the separated laminar boundary layer are computed by the method of Computational Aero-Acoustics (CAA), and the effects of Reynolds number, Mach number and adverse pressure gradient strength are examined.

• 한국전자파학회논문지
• /
• 제9권4호
• /
• pp.552-559
• /
• 1998

Conducted noise in power line contains both the common mode(CM) and differential mode(DM) noise. These two modes of noise are caused by different noise sources and paths. Therefore, CM/DM noise must be deal with individually in EMI filter. In this paper the technique to separate power line noise is presented, which can be used to measure both the CM and the DM noise from total generated noise. Also, noise-separator is designed and experimental results showed 30 dB above of separation performance in 10 kHz~10 MHz.

• 한국음향학회지
• /
• 제35권2호
• /
• pp.110-117
• /
• 2016

수동소나 시스템에서는 함정의 소음원에서 발생하는 방사 소음을 분석하여 표적을 탐지 및 식별한다. 소나의 탐지 범위 안에 다수의 소음원이 존재하면 신호를 분석할 때 각 소음원에서 나오는 성분들이 혼합되어 각각의 소음원을 규명하기가 어렵다. 이를 해결하기 위해 일반적으로는 배열 센서를 이용한 빔을 형성하여 소음원의 신호를 공간적으로 분리하는 기법이 사용되지만 환경에 따라 여전히 어려운 점이 있다. 본 연구에서는 수동소나 표적신호를 분리하기 위한 새로운 방법으로 주파수영역 독립성분분석(FDICA: Frequency Domain Independent Component Analysis)을 적용하고, 혼합된 표적신호를 분리하는 모의실험을 수행하여 그 타당성을 검증하였다. 표적신호 합성을 위한 특징 정보로는 기계류 토널 성분 및 프로펠러 성분을 사용하였고, 분리 전 후의 결과를 LOFAR(Low Frequency Analysis and Recording), DEMON(Detection Envelope Modulation On Noise) 분석을 통해 비교하였다.

• 대한음성학회지:말소리
• /
• 제48호
• /
• pp.81-91
• /
• 2003

In this paper, we propose a speech recognition method based on independent component analysis (ICA) and spectral enhancement techniques. While ICA tris to separate speech signal from noisy speech using multiple channels, some noise remains by its algorithmic limitations. Spectral enhancement techniques can compensate for lack of ICA's signal separation ability. From the speech recognition experiments with instantaneous and convolved mixing environments, we show that the proposed approach gives much improved recognition accuracies than conventional methods.

• 대한음성학회:학술대회논문집
• /
• 대한음성학회 2003년도 10월 학술대회지
• /
• pp.97-100
• /
• 2003

This paper addresses a method of convolutive source separation that based on SEONS (Second Order Nonstationary Source Separation) [1] that was originally developed for blind separation of instantaneous mixtures using nonstationarity. In order to tackle this problem, we transform the convolutive BSS problem into multiple short-term instantaneous problems in the frequency domain and separated the instantaneous mixtures in every frequency bin. Moreover, we also employ a H infinity filtering technique in order to reduce the sensor noise effect. Numerical experiments are provided to demonstrate the effectiveness of the proposed approach and compare its performances with existing methods.

• Transactions on Control, Automation and Systems Engineering
• /
• 제2권1호
• /
• pp.7-12
• /
• 2000

A new speech enhancement algorithm for speech corrupted by slowly varying additive colored noise is suggested based on a state-space signal model. Due to the FIR structure and the unimportance of long-term past information, the receding horizon (RH) FIR filter known to be a best linear unbiased estimation (BLUE) filter is utilized in order to obtain noise-suppressed speech signal. As a special case of the colored noise problem, the suggested approach is generalized to perform the single blind signal separation of two speech signals. It is shown that the exact speech signal is obtained when an incoming speech signal is noise-free.

• 한국자동차공학회논문집
• /
• 제5권3호
• /
• pp.35-46
• /
• 1997

In order to satisfy customer's requirements of ride comfort and high performance, it is necessary for designers to fully understand vehicle aerodynamics and wind noise of newly produced cars because characteristics of flow and wind noise are heavily dependent on each other. In this study numerical and experimental study have been carried out to analyse the effect of flow characteristics at around of rear view mirror on wind noise and soiling on the front S/W. As a result, it's found that the spiral flow mear the front pillar is weakened and spreaded because rear view mirror obstructs the flow. It is also shown that there is abrupt change of gradient of separa- tion line, separation area, intensity of spiral flow and turbulent kinetic energy with varying shape of neck and housing of rear view mirror.

• 한국운동역학회지
• /
• 제13권1호
• /
• pp.23-38
• /
• 2003

The purpose of this study is to develop an improved marker auto-identification algorithm for reduce of data processing time through improve the efficiency of noise elimination and marker separation. The maker auto-identification algorithm was programming named KUMAS used Delphi language. For the study, various experiments were conducted for the verification of KUMAS. and compared two systems of established with the KUMAS. Four different motions - cycling, gait, rotation, and pendulum -, were selected and tested. Motions were filmed 30Hz frames rate per second. ${\chi}^2$ used for statistical analysis. Significant level were ${\alpha}=.05$. The test results were as follow. 1. Increased the success ratio of marker auto-identification. 2. The efficiency of marker auto-identification was remarkably improved through marker separation, noise elimination. 3. The marker auto-identification ability was improved in 2D-image plane include the 3D motion. 4. Significant different were found between KUMAS and B-SYS(established system) with non-input the artificial noise frames, input the artificial noise frames and total frames.

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 2013년도 춘계학술대회 논문집
• /
• pp.520-521
• /
• 2013