• The Journal of the Acoustical Society of Korea
• /
• v.17 no.1
• /
• pp.55-59
• /
• 1998

본 논문에서는 잡음에 강한 음성 인식기를 위한 음성의 특징 추출에 관해서 살펴 보았다. 지금까지의 음성 인식기는 조용한 실험실 환경하에서 학습이 이루어지나 실제 테스 트는 여러 가지 환경에서 이루어지므로, 이러한 환경 변화에 따라 음성인식 시스템의 성능 이 감소함을 보여왔다. 이를 보완하기 위해 여러 가지 연구가 진행되고 있으나 본 연구에서 는 음성의 특징 추출 부분에서 순시적인 신호대 잡음비 예측과 잡음에 강한(RASTA)처리를 하므로써 인식율을 향상시켰다.

• Journal of Biomedical Engineering Research
• /
• v.20 no.3
• /
• pp.299-305
• /
• 1999

In thls paper, the representative beal textraction algorIthms for the diagnostic parameter extraction in noisy signal were compared. We used the avernge, median, mode, and trmmed mean to calculale the central tendency. In our experimenl, we have restricted to four kinds of noises -EMG noise, 60Hz powerline inlerference, ahrupl baseline shift, and baselme drift due to respimtion-which were commonly occurred in ECG mgnal, then we have calculated signal-to-noise ratios(SNRs) for the ECG corrupted with each noise and all noises together. As the result of this paper, we have proved that the average method has super lor performance than the others in the ECG corrupted wilh EMG noise. When the signal mcludes extreme value such as abrupt baseline shIft, the median, mode, trimmed mean methods have supenor performance in the SNR ratios. Especially when the ECG corrupted with baseline drift due to respirallon, the trimmed mean method was most efficient because ST level change was 0 V.

• Journal of Korea Multimedia Society
• /
• v.4 no.1
• /
• pp.46-55
• /
• 2001

In this paper, a new voice activity detection method estimating SNR of enhanced speech with extended spectral subtraction (ESS) is proposed. Voice activity detection is performed by putting an second Wiener filter behind an Wiener filter used in the ESS to estimate speech and noise power of output signal of first Wiener filter. The proposed voice activity detection method does not require many computational loads and performs well under severe input SNR. Boll's spectral substraction algorithm with proposed voice activity detection was compared to ESS under several noise environment having different time-frequency distributions. During speech and non-speech activity, performance of Boll's spectral substraction algorithm with proposed voice activity detection is superior to that of ESS.

• The Journal of the Acoustical Society of Korea
• /
• v.11 no.1E
• /
• pp.15-22
• /
• 1992

잡음이 있는 환경하에서는 음성 인식의 성능이 현저하게 떨어지게 된다. 본 논문에서는 이렇나 잡음의 영향에 강한 거리척도를 제안하고자 한다. 우리는 잡음이 더해진 음성신호의 특징벡터를 깨끗한 음성신호의 특징벡터가 FIR 시스템을 거쳐 변형된 것이라고 가정한다. 여기서 FIR 시스템은 잡음의 영 향을 모델링한 것이라고 할 수 있다. 미지의 FIR 시스템 계수잡음의 영향을 모델링한 것이라고 할 수 있다. 미지의 FIR 시스템계수들은 RLS 적응 알고리즘을 이용하여 구한다. 제안된 거리척도는 적응 여파 기의 예측 오차에 관한 식으로 표시되어진다. 여러 가지 적응 여파기의 구조중 단일 채널 일차 FIR 구 조가 가장 좋은 음성 인식 성능을 보이며, 이 경우 효과적인 거리척도 알고리즘을 구할 수 있다. 여러 가지 신호대 잡음비에 관하여 화자독립 격리단어 인식 실험을 DTW 알고리즘을 이용하여 수행하여 본 결과 제안된 거리척도가 거의 모든 신호대 잡음비에 대하여 우수한 성능을 보였다.

• Proceedings of the IEEK Conference
• /
• 2000.09a
• /
• pp.225-228
• /
• 2000

본 논문에서는 2차의 격자 IIR 노치 필터를 이용하여DS-CDMA 시스템에서 협대역 간섭신호를 제거한다. 본 노치필터는 구조가 간단하고 다양한 간섭 신호 모델에 따라 필터의 계수를 조절할 수 있어 간섭신호의 효과적인 제거가 가능하다. 노치필터의 앞단에는 주파수 추정부를 평행하게 구성하여 간섭신호 주파수의 위치와 파워를 검출한다. 제안된 적응 필터링 알고리듬은 이러한 정보를 이용하여 노치필터의 주파수를 조절하고 간섭 신호의 파워와 대역폭에 따라 노치의 넓이와 깊이를 조절한다. 즉, IIR 노치 필터의 특성을 이론적으로 분석하여 출력 신호대 잡음비를 입력과 필터의 파라미터에 관한 수식으로 유도하였고 이를 이용하여 주어진 입력에 따른 최적의 필터 파라미터를 구하여 적용시키는 것이다. 입력의 간섭 주파수와 파워를 검출하는 방법에는 여 러가지가 있지만 본 논문에서는 비교적 간단한 IIR ALE[6]를 사용하였다. 제안된 알고리듬을 사용한 결과 신호대 잡음비와 에러율에 있어서 FIR 필터를 사용한 기존의 방법[4]에 비하여 좋은 성능을 보임을 알 수 있다.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.3C
• /
• pp.323-330
• /
• 2007

Performance indicators such as output SNR, SIR, SINR for rectangular smart antennas in CDMA basestations have been derived. Simulations have been carried out to find the rectangular smart antenna performance while varying the input SNR, number of antenna elements, and the interferers' spatial distributions. Simplified Conjugate Gradient Method was chosen as the underlying beam forming algorithm. It has been shown that the performance of a rectangular smart antenna is similar to that of the linear one having the same number of elements when the interferers are randomly distributed over the whole azimuth angle range.

• Geophysics and Geophysical Exploration
• /
• v.5 no.4
• /
• pp.250-256
• /
• 2002

Resistivity tomography has become an important tool to image underground resistivity distribution. This method has been widely applied to site investigation for engineering and environmental purpose. In resistivity tomography, various electrode arrays can be used and each array has both merits and demerits. For example, the pole-pole array has high signal to noise ratio (S/N ratio), but its resolution is too low. The dipole-dipole array has low S/N ratio, but its resolution is very high. The Pole-dipole may has intermediate Snf ratio and resolution. The modified Pole-dipole array, recently proposed, shows reasonable S/N ratio and resolution, which are comparable to the pole-dipole array. These electrode arrays except the pole-pole array, however, have the problem that the apparent resistivity can diverge at some special electrode Positions. Also, the Pole-Pole array may not reflect the doe resistivity of an anomalous body. In this study, we propose a new electrode array, mixed array, where pole-dipole and modified pole-dipole ways are selectively used with the relative positions of current and potential electrodes. The mixed array has the same level of S/N ratio and resolution as the pole-dipole array and the apparent resistivity does not diverge in the receiver hole. Furthermore, the apparent resistivity using the array can reflect the true resistivity of the anomalous body.

• The Journal of the Acoustical Society of Korea
• /
• v.22 no.8
• /
• pp.670-679
• /
• 2003

This paper deals with application of wavelet transform, which is known to be good for time-frequency analysis, in order to detect the underwater transient signals embedded in ambient noise. A new detector of acoustic transient signals is presented. It combines two detection tools: wavelet analysis and higher-order statistics. Using both techniques, the detection of the transient signal is possible in low signal to noise ratio condition. The proposed algorithm uses the wavelet transform of a partition of the signal on frequency domain, and then higher-order statistics tests the Gaussian nature of the segments.

• Journal of Broadcast Engineering
• /
• v.17 no.6
• /
• pp.1061-1068
• /
• 2012

This paper deals with a noise reduction algorithm which uses the binary masking approach in the time-frequency domain to improve speech intelligibility. In the binary masking approach, the noise-corrupted speech is decomposed into time-frequency units. Noise-dominant time-frequency units are removed by setting the corresponding binary masks as "0"s and target-dominant units are retained untouched by assigning mask "1"s. We propose a binary mask estimation by comparing the local signal-to-noise ratio (SNR) to a threshold. The local SNR is estimated by a training-based approach. An optimal threshold is proposed, which is obtained from observing the distribution of the training database. The proposed method is evaluated by normal-hearing subjects and the intelligibility scores are computed by counting the number of words correctly recognized.

• Investigative Magnetic Resonance Imaging
• /
• v.8 no.1
• /
• pp.42-49
• /
• 2004

Purpose : The quantization noise in magnetic resonance imaging (MRI) systems is analyzed. The signal-to-quantization noise ratio (SQNR) in the reconstructed image is derived from the level of quantization in the signal in spatial frequency domain. Based on the derived formula, the SQNRs in various main magnetic fields with different receiver systems are evaluated. From the evaluation, the quantization noise could be a major noise source determining overall system signal-to-noise ratio (SNR) in high field MRI system. A few methods to reduce the quantization noise are suggested. Materials and methods : In Fourier imaging methods, spin density distribution is encoded by phase and frequency encoding gradients in such a way that it becomes a distribution in the spatial frequency domain. Thus the quantization noise in the spatial frequency domain is expressed in terms of the SQNR in the reconstructed image. The validity of the derived formula is confirmed by experiments and computer simulation. Results : Using the derived formula, the SQNRs in various main magnetic fields with various receiver systems are evaluated. Since the quantization noise is proportional to the signal amplitude, yet it cannot be reduced by simple signal averaging, it could be a serious problem in high field imaging. In many receiver systems employing analog-to-digital converters (ADC) of 16 bits/sample, the quantization noise could be a major noise source limiting overall system SNR, especially in a high field imaging. Conclusion : The field strength of MRI system keeps going higher for functional imaging and spectroscopy. In high field MRI system, signal amplitude becomes larger with more susceptibility effect and wider spectral separation. Since the quantization noise is proportional to the signal amplitude, if the conversion bits of the ADCs in the receiver system are not large enough, the increase of signal amplitude may not be fully utilized for the SNR enhancement due to the increase of the quantization noise. Evaluation of the SQNR for various systems using the formula shows that the quantization noise could be a major noise source limiting overall system SNR, especially in three dimensional imaging in a high field imaging. Oversampling and off-center sampling would be an alternative solution to reduce the quantization noise without replacement of the receiver system.