• Phonetics and Speech Sciences
• /
• v.7 no.4
• /
• pp.17-25
• /
• 2015

For blind source separation of convolutive mixtures, FDICA(Frequency Domain Independent Component Analysis) algorithms are generally used. Since FDICA algorithm such as Sawada FDICA, IVA(Independent Vector Analysis) works on the frequency bin basis with a natural gradient descent method, it takes much time to converge. In this paper, we propose a new method to improve convergence speed in FDICA algorithm. The proposed method reduces the number of iteration drastically in the process of natural gradient descent method by applying a weighted inner product constraint of unmixing matrix. Experimental results have shown that the proposed method achieved large improvement of convergence speed without degrading the separation performance of the baseline algorithms.

• Wind and Structures
• /
• v.37 no.6
• /
• pp.445-460
• /
• 2023

Accurate estimation of modal parameters (i.e., natural frequency, damping ratio) of tall buildings is of great importance to their structural design, structural health monitoring, vibration control, and state assessment. Based on the combination of variational mode decomposition, smoothed discrete energy separation algorithm-1, and Half-cycle energy operator (VMD-SH), this paper presents a method for structural modal parameter estimation. The variational mode decomposition is proved to be effective and reliable for decomposing the mixed-signal with low frequencies and damping ratios, and the validity of both smoothed discrete energy separation algorithm-1 and Half-cycle energy operator in the modal identification of a single modal system is verified. By incorporating these techniques, the VMD-SH method is able to accurately identify and extract the various modes present in a signal, providing improved insights into its underlying structure and behavior. Subsequently, a numerical study of a four-story frame structure is conducted using the Newmark-β method, and it is found that the relative errors of natural frequency and damping ratio estimated by the presented method are much smaller than those by traditional methods, validating the effectiveness and accuracy of the combined method for the modal identification of the multi-modal system. Furthermore, the presented method is employed to estimate modal parameters of a full-scale tall building utilizing acceleration responses. The identified results verify the applicability and accuracy of the presented VMD-SH method in field measurements. The study demonstrates the effectiveness and robustness of the proposed VMD-SH method in accurately estimating modal parameters of tall buildings from acceleration response data.

• 제어로봇시스템학회:학술대회논문집
• /
• 1987.10b
• /
• pp.83-86
• /
• 1987

A vision algorithm is presented for the separation of near objects from distant objects. In the algorithm, a difference field of a stereo pair of images is computed to obtain the range information and the median filter is used for the suppression of distant objects. The objects within a given distance is segmented by thresholding the gray scale cross-section of the median filtered difference field. The experiment is performed in a laboratory setting.

• ETRI Journal
• /
• v.39 no.1
• /
• 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.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.3855-3867
• /
• 2022

Name separation (splitting full names into surnames and given names) is not a tedious task in a multiethnic country because the procedure for splitting surnames and given names is ethnicity-specific. Malaysia has multiple main ethnic groups; therefore, separating Malaysian full names into surnames and given names proves a challenge. In this study, we develop a two-phase framework for Malaysian name separation using deep learning. In the initial phase, we predict the ethnicity of full names. We propose a recurrent neural network with long short-term memory network-based model with character embeddings for prediction. Based on the predicted ethnicity, we use a rule-based algorithm for splitting full names into surnames and given names in the second phase. We evaluate the performance of the proposed model against various machine learning models and demonstrate that it outperforms them by an average of 9%. Moreover, transfer learning and fine-tuning of the proposed model with an additional dataset results in an improvement of up to 7% on average.

• The Journal of the Acoustical Society of Korea
• /
• v.35 no.1
• /
• pp.49-54
• /
• 2016

In this paper, we propose a vocal separation method using weighted ${\beta}$-order minimum mean wquare error estimation (WbE) based on kernel back-fitting algorithm. In spoken speech enhancement, it is well-known that the WbE outperforms the existing Bayesian estimators such as the minimum mean square error (MMSE) of the short-time spectral amplitude (STSA) and the MMSE of the logarithm of the STSA (LSA), in terms of both objective and subjective measures. In the proposed method, WbE is applied to a basic iterative kernel back-fitting algorithm for improving the vocal separation performance from monaural music signal. The experimental results show that the proposed method achieves better separation performance than other existing methods.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.1A
• /
• pp.73-81
• /
• 2011

IEEE 802.11 PHY and MAC layer provide multiple channels and data rates. To improve the performance of IEEE 802.11 multi-radio ad hoc networks, it is required to utilize available channels and data rates efficiently. However, in IEEE 802.11 multi-rate networks, the rate anomaly (RA) problem occurs that the network performance is severely degraded as low-rate links affect high-rate links. Hence, in this paper, we propose a model-based rate separation (MRS) algorithm that uses multiple channels to separate different data rate links so that the RA problem is mitigated. MRS algorithm utilizes an existing throughput model that estimates the throughput of IEEE 802.11 single-hop networks to separate low-rate links and high-rate links. Through simulations, we demonstrate that the MRS algorithm shows improved network performance compared with existing algorithms in multi-radio ad hoc networks.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.11
• /
• pp.87-97
• /
• 2013

In the RFID system, multiple tags respond in the process of identifying multiple tags in the reader's interrogation zone, resulting in collisions. Tag collision occurs when two or more tags respond to one reader, so that the reader cannot identify any tags. These collisions make it hard for the reader to identify all tags within the interrogation zone and delays the identifying time. In some cases, the reader cannot identify any tags. The reader needs the anti-collision algorithm which can quickly identify all the tags in the interrogation zone. The proposed algorithm efficiently divides tag groups through an efficient separation to respond, preventing collisions. Moreover, the proposed algorithm identifies tags without checking all the bits in the tags. The prediction with efficient separation reduces the number of the requests from the reader.

• Journal of applied mathematics & informatics
• /
• v.26 no.5_6
• /
• pp.1149-1159
• /
• 2008

Blind source extraction (BSE) is a special class of blind source separation (BSS) methods, which only extracts one or a subset of the sources at a time. Based on the time delay of the desired signal, a simple but important extraction algorithm (simplified " BC algorithm")was presented by Barros and Cichocki. However, the performance of this method is not satisfying in some cases for which it only carries out the constrained minimization of the mean squared error. To overcome these drawbacks, ICA with reference (ICA-R) based approach, which considers the higher-order statistics of sources, is added as the second stage for further source extraction. Specifically, BC algorithm is exploited to roughly extract the desired signal. Then the extracted signal in the first stage, as the reference signal of ICA-R method, is further used to extract the desired sources as cleanly as possible. Simulations on synthetic data and real-world data show its validity and usefulness.

• The Journal of the Acoustical Society of Korea
• /
• v.27 no.8
• /
• pp.418-426
• /
• 2008

In this paper, we propose an algorithm for the frequency channel segmentation using peaks and valleys in spectrogram. The frequency channel segments means that local groups of channels in frequency domain that could be arisen from the same sound source. The proposed algorithm is based on the smoothed spectrum of the input sound. Peaks and valleys in the smoothed spectrum are used to determine centers and boundaries of segments, respectively. To evaluate a suitableness of the proposed segmentation algorithm before that the grouping stage is applied, we compare the synthesized results using ideal mask with that of proposed algorithm. Simulations are performed with mixed speech signals with narrow band noises, wide band noises and other speech signals.