• Speech Sciences
• /
• 제11권4호
• /
• pp.53-66
• /
• 2004

The nonlinear speech enhancement method with interactive parallel-extended Kalman filter is applied to speech contaminated by additive white noise. To represent the nonlinear and nonstationary nature of speech. we assume that speech is the output of a nonlinear prediction HMM (NPHMM) combining both neural network and HMM. The NPHMM is a nonlinear autoregressive process whose time-varying parameters are controlled by a hidden Markov chain. The simulation results shows that the proposed method offers better performance gains relative to the previous results [6] with slightly increased complexity.

• The Journal of the Acoustical Society of Korea
• /
• 제25권4E호
• /
• pp.154-158
• /
• 2006

In this paper, we suggest a noise estimation method for speech enhancement in nonstationary noisy environments. The proposed method consists of the following two main processes. First, in order to receive fewer affect of variable signals, a best fitting regression line is used, which is obtained by applying a least squares method to coefficient magnitudes in a node with a uniform wavelet packet transform. Next, in order to update the noise estimation efficiently, a differential forgetting factor and a correlation coefficient per subband are used, where subband is employed for applying the weighted value according to the change of signals. In particular, this method has the ability to update the noise estimation by using the estimated noise at the previous frame only, without utilizing the statistical information of long past frames and explicit nonspeech frames by voice activity detector. In objective assessments, it was observed that the performance of the proposed method was better than that of the compared (minima controlled recursive averaging, weighted average) methods. Furthermore, the method showed a reliable result even at low SNR.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• 제3권1호
• /
• pp.83-95
• /
• 1999

Wavelet transform has localization for time or frequency. It is useful to analyze a nonstationary signal. Basic function on wavelet transform is generated dilating and translating the original wavelet(mother wavelet). In this paper, time-frequency analysis method using linear combination wavelet transform is proposed. And it is applied to diagnostic monitoring system using the proposed linear combination wavelet transform. The stationary and nonstationary signal is used linear chirp signal, fan noise signal, a sinusoid signal from revolution body, electronic signal. Transform applied to signal analysis use fast Fourier transform (FFT), Daubechies, Haar and proposed linear combination method. The result of time-frequency analysis using linear combination wavelet transform is suited for portraying nonstationary time signal as well as stationary signal. Also the diagnostic monitoring system carry out the effective the signal analysis.

• Journal of the Institute of Electronics and Information Engineers
• /
• 제51권9호
• /
• pp.165-170
• /
• 2014

Speech enhancement techniques that remove surrounding noise are stressed to preprocessor of speech recognition. Among the various speech enhancement techniques, Codebook-based Speech Enhancement (CBSE) operates efficiently in non-stationary noise environments. But, CBSE has some problems that inaccurate gains can be estimated if mismatch occur between input noisy signal and trained speech/noise codevectors. In this paper, the Normalized Weighting Factor (NWF) is calculated by long-term noise estimation algorithm based on Signal-to-Noise Ratio, compensated to the conventional inaccurate gains. The proposed CBSE shows better performance than conventional CBSE.

• Journal of Communications and Networks
• /
• 제16권1호
• /
• pp.92-97
• /
• 2014

A novel method for extracting frequency slippage signal from radar full pulse sequence is presented. For the radar full pulse sequence received by radar interception receiver, radio frequency (RF) and time of arrival (TOA) of all pulses constitute a two-dimensional information sequence. In a complex and intensive electromagnetic environment, the TOA of pulses is distributed unevenly, randomly, and in a nonstationary manner, preventing existing methods from directly analyzing such time series and effectively extracting certain signal features. This work applies Gaussian noise insertion and structure function to the TOA-RF information sequence respectively such that the equalization of time intervals and correlation processing are accomplished. The components with different frequencies in structure function series are separated using empirical mode decomposition. Additionally, a chaos detection model based on the Duffing equation is introduced to determine the useful component and extract the changing features of RF. Experimental results indicate that the proposed methodology can successfully extract the slippage signal effectively in the case that multiple radar pulse sequences overlap.

• The Journal of the Acoustical Society of Korea
• /
• 제14권3호
• /
• pp.28-36
• /
• 1995

In this paper, a double loop control model using leaky delay LMS algorithm are proposed for active noise control. The proposed double loop control model estimates the loudspeaker characteristic and the error path transfer function with on-line using only gain and acoustic time delay to reduce computation burden. The control of error signal through double loop control scheme makes the more robust cntrol system. The input signal of filter to estimate acoustic time delay is used difference between input signal of input microphone and adaptive filter output. And also, in nonstationary environments, the leaky delay LMS algorithm is employed to counteract parameter drift of delay LMS algorithm. For practical noise signal, the proposed double loop control model reduces noise level about 12.9 dB.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• 제6권4호
• /
• pp.639-644
• /
• 2002

Nonlinear filters which are utilized rank-order information and temporal-order information, have many proposed, in order to restore nonstationary signals which are corrupted by additive noise. In this paper, we propose a data-dependent LWOS filter whose coefficients change based on local statistics. LWOS(Linear Combination of Weighted Order Statistics) filters［1］which also utilized two informations, and have properties of efficient impulsive and nonimpulsive noise attenuation and sufficiently details and edges preservation. DD-LWOS filters can remove non-impulsive oises while preserving signal details. DD-LWOS2 filter gets more better performance than DD-LWOS filter when input image corrupted by additive noise which includes Impulsive noise components.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 한국소음진동공학회 2005년도 추계학술대회논문집
• /
• pp.551-554
• /
• 2005

The method of the reduction of the automotive induction noise can be classified by the method of passive control and the method of active control. However, the passive control method has a demerit to reduce the effect of noise reduction at low frequency (below 500Hz) range and to be limited by a space of the engine room. Whereas, the active control method can overcome the demerit of passive control method. The algorithm of active control is mostly used the LMS (Least-Mean-Square) algorithm because the LMS algorithm can easily obtain the complex transfer function in real-time. Especially, When the Filtered-X LMS (FXLMS) algorithm is applied to an ANC system. However, the convergence performance of LMS algorithm goes bad when the FXLMS algorithm is applied to an active control of the induction noise under rapidly accelerated driving conditions. Thus Normalized FXLMS algorithm was developed to improve the control performance under the rapid acceleration. The advantage of Normalized FXLMS algorithm is that the step size is no longer constant. Instead, it varies with time. But there is one additional practical difficulty that can arise when a nonstationary input is used. If the input is zero for consecutive samples, then the step size becomes unbounded. So, in order to solve this problem. the Correlation FXLMS algorithm was developed. The Correlation FXLMS algorithm is realized by using an estimate of the cross correlation between the adaptation error and the filtered input signal to control the step size. In this paper, the performance of the Correlation FXLMS Is presented in comparison with that of the other FXLMS algorithms based on computer simulations.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• 제25권2호
• /
• pp.202-207
• /
• 2021

One of the most important tasks in image processing is noise filtering. Noise removal in image is a difficult task due to many reasons such as nonstationary sequences and corrupted by various types of noise. Human's visual perception is heavily based on the edge information. Thus, noise filtering must preserve edges. To remove the noise, we usually use the square-shaped median filter. They possess mathematical simplicity but have the disadvantages that blur the edges. In this paper we consider a new technique for image restoration using a weighted octagonal median filter. The technique consists of simple hypothesis test for edge detection, and we use the weighted octagonal-shaped moving window. The new technique is applied to noise corrupted image and experimental results are compared to the results of the square-shaped median filter and the cross-shaped median filter.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• 제19권1호
• /
• pp.61-68
• /
• 2015

Noise reduction is necessary to compensate for the degradation of recognition performance by various types of noises. Among many noise reduction techniques using microphone array, generalized sidelobe canceller (GSC) has been widely applied to reduce nonstationary noises. The performance of GSC is directly affected by its adaptation mode controller (AMC). That is, accurate target speech detection is essential to guarantee the sufficient noise reduction in pure noise intervals and the less distortion in target speech intervals. Thus, this paper proposes an improved AMC design technique in which the power ratio of the output of fixed beamforming to that of blocking matrix is calculated frequency bandwise and probabilistically modeled by mixture Gaussians for each class. Experimental results show that the proposed algorithm outperforms conventional AMCs in receiver operating curves (ROC) and output SNRs.