• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.2
• /
• pp.15-23
• /
• 2017

The person Re-identification is the most challenging part of computer vision due to the significant changes in human pose and background clutter with occlusions. The picture from non-overlapping cameras enhance the difficulty to distinguish some person from the other. To reach a better performance match, most methods use feature selection and distance metrics separately to get discriminative representations and proper distance to describe the similarity between person and kind of ignoring some significant features. This situation has encouraged us to consider a novel method to deal with this problem. In this paper, we proposed an enhanced recurrent neural network with three-tier hierarchical network for person re-identification. Specifically, the proposed recurrent neural network (RNN) model contain an iterative expectation maximum (EM) algorithm and three-tier Hierarchical network to jointly learn both the discriminative features and metrics distance. The iterative EM algorithm can fully use of the feature extraction ability of convolutional neural network (CNN) which is in series before the RNN. By unsupervised learning, the EM framework can change the labels of the patches and train larger datasets. Through the three-tier hierarchical network, the convolutional neural network, recurrent network and pooling layer can jointly be a feature extractor to better train the network. The experimental result shows that comparing with other researchers' approaches in this field, this method also can get a competitive accuracy. The influence of different component of this method will be analyzed and evaluated in the future research.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.452-459
• /
• 2021

Estimating the mutual time delay between two acoustic sensors is used in various fields such as tracking and estimating the location of a target in room acoustics and sonar. In the time delay estimation methods, there are a non-parametric method, such as Generalized Cross Correlation (GCC), and a parametric method based on system identification. In this paper, we propose a time delay estimation method based on the parametric method. In particular, we propose a method that considers the noise in each receiving acoustic sensor. Simulation confirms that the proposed algorithm is superior to the existing generalized cross-correlation and adaptive eigenvalue analysis methods in white noise and reverberation environments.

• Journal of Communications and Networks
• /
• v.5 no.2
• /
• pp.157-166
• /
• 2003

This paper deals with the problem of channel identification for Single Input Multiple Output (SIMO) slow fading channels using clustering algorithms. Due to the intrinsic memory of the discrete-time model of the channel, over short observation periods, the received data vectors of the SIMO model are spread in clusters because of the AWGN noise. Each cluster is practically centered around the ideal channel output labels without noise and the noisy received vectors are distributed according to a multivariate Gaussian distribution. Starting from the Markov SIMO channel model, simultaneous maximum ikelihood estimation of the input vector and the channel coefficients reduce to one of obtaining the values of this pair that minimizes the sum of the Euclidean norms between the received and the estimated output vectors. Viterbi algorithm can be used for this purpose provided the trellis diagram of the Markov model can be labeled with the noiseless channel outputs. The problem of identification of the ideal channel outputs, which is the focus of this paper, is then equivalent to designing a Vector Quantizer (VQ) from a training set corresponding to the observed noisy channel outputs. The Linde-Buzo-Gray (LBG)-type clustering algorithms [1] could be used to obtain the noiseless channel output labels from the noisy received vectors. One problem with the use of such algorithms for blind time-varying channel identification is the codebook initialization. This paper looks at two critical issues with regards to the use of VQ for channel identification. The first has to deal with the applicability of this technique in general; we present theoretical results for the conditions under which the technique may be applicable. The second aims at overcoming the codebook initialization problem by proposing a novel approach which attempts to make the first phase of the channel estimation faster than the classical codebook initialization methods. Sample simulation results are provided confirming the effectiveness of the proposed initialization technique.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.45 no.12
• /
• pp.105-113
• /
• 2008

In Ubiquitous Society, accurate Location Calculation of user's device is required to achieve the need of users. As the location calculation is processed by ranging between transceivers, if some obstacles exist between transceivers, NLoS(Non-line-of-Sight) components of received signal increase along with the reduction of LoS(Line-of-Sight) components. Therefore the location calculation error will increase due to the NLoS effect. The conventional location calculation algorithm has the original ranging error because there is no transformation of ranging information which degrades the ranging accuracy. The Iterative Calculation method which minimizes the location calculation error relys on accurately identifying NLoS or LoS condition of the tested channel. We employ Kurtosis, Mean Excess Delay and RMS Delay spread of the received signal to identify whether the tested channel is LoS or NLoS firstly. Thereafter, to minimize location calculation error, the proposed Iterative Calculation method iteratively select random range and finds the averaged target location which has high probability. The simulation results confirm the enhancement of the proposed method.

• Smart Structures and Systems
• /
• v.6 no.5_6
• /
• pp.579-593
• /
• 2010

Low-power radio frequency (RF) chip transceiver technology and the associated structural health monitoring platforms have matured recently to enable high-rate, lossless transmission of measurement data across large-scale sensor networks. The intrinsic value of these advanced capabilities is the allowance for high-quality, rapid operational modal analysis of in-service structures using distributed accelerometers to experimentally characterize the dynamic response. From the analysis afforded through these dynamic data sets, structural identification techniques can then be utilized to develop a well calibrated finite element (FE) model of the structure for baseline development, extended analytical structural evaluation, and load response assessment. This paper presents a case study in which operational modal analysis is performed on a three-span prestressed reinforced concrete bridge using a wireless sensor network. The low-power wireless platform deployed supported a high-rate, lossless transmission protocol enabling real-time remote acquisition of the vibration response as recorded by twenty-nine accelerometers at a 256 Sps sampling rate. Several instrumentation layouts were utilized to assess the global multi-span response using a stationary sensor array as well as the spatially refined response of a single span using roving sensors and reference-based techniques. Subsequent structural identification using FE modeling and iterative updating through comparison with the experimental analysis is then documented to demonstrate the inherent value in dynamic response measurement across structural systems using high-rate wireless sensor networks.

• Smart Structures and Systems
• /
• v.21 no.4
• /
• pp.449-467
• /
• 2018

In this paper, a theoretical and numerical study on bridge simultaneous damage detection procedure for identifying both the system parameters and input excitation mass, are presented. This method is called 'Adjoint Variable Method' which is an iterative gradient-based model updating method based on the dynamic response sensitivity. The main advantage of proposed method is inclusion of an analytical method to augment the accuracy and speed of the solution. Moving mass is a model which takes into account the inertia effects of the vehicle. This interaction model is a time varying system and proposed method is capable of detecting damage in this variable system. Robustness of proposed method is illustrated by correctly detection of the location and extension of predetermined single, multiple and random damages in all ranges of speed and mass ratio of moving vehicle. A comparison study of common sensitivity and proposed method confirms its efficiency and performance improvement in sensitivity-based damage detection methods. Various sources of errors including the effects of measurement noise and initial assumption error in stability of method are also discussed.

• Wind and Structures
• /
• v.24 no.4
• /
• pp.351-365
• /
• 2017

The minimization method is widely used to predict the dynamic characteristics of a system. Generally, data recorded by experiment (for example displacement) tends to contain noise, and the error in the properties of the system is proportional to the noise level (NL). In addition, the accuracy of the results depends on various factors such as the signal character, filtering method or cut off frequency. In particular, coupled terms in multimode systems show larger differences compared to the true value when measured in an environment with a high NL. The iterative least square (ILS) method was proposed to reduce these errors that occur under a high NL, and has been verified in previous research. However, the ILS method might be sensitive to the signal processing, including the determination of cutoff frequency. This paper focused on improving the accuracy of the ILS method, and proposed the modified ILS (MILS) method, which differs from the ILS method by the addition of a new calculation process based on correlation coefficients for each degree of freedom. Comparing the results of these systems with those of a numerical simulation revealed that both ILS and the proposed MILS method provided good prediction of the dynamic properties of the system under investigation (in this case, the damping ratio and damped frequency). Moreover, the proposed MILS method provided even better prediction results for the coupling terms of stiffness and damping coefficient matrix.

• Smart Structures and Systems
• /
• v.17 no.2
• /
• pp.167-194
• /
• 2016

This paper investigates the Synergetics based Damage Detection Method (SDDM) for frame structures by using surface-bonded PZT (Lead Zirconate Titanate) patches. After analyzing the mechanism of pattern recognition from Synergetics, the operating framework with cooperation-competition-update process of SDDM was proposed. First, the dynamic identification equation of structural conditions was established and the adjoint vector (AV) set of original vector (OV) set was obtained by Generalized Inverse Matrix (GIM).Then, the order parameter equation and its evolution process were deduced through the strict mathematics ratiocination. Moreover, in order to complete online structural condition update feature, the iterative update algorithm was presented. Subsequently, the pathway in which SDDM was realized through the modified Synergetic Neural Network (SNN) was introduced and its assessment indices were confirmed. Finally, the experimental platform with a two-story frame structure was set up. The performances of the proposed methodology were tested for damage identifications by loosening various screw nuts group scenarios. The experiments were conducted in different damage degrees, the disturbance environment and the noisy environment, respectively. The results show the feasibility of SDDM using piezoceramic sensors and actuators, and demonstrate a strong ability of anti-disturbance and anti-noise in frame structure applications. This proposed approach can be extended to the similar structures for damage identification.

• ETRI Journal
• /
• v.36 no.1
• /
• pp.175-178
• /
• 2014

We introduce the Energy-Based Blind Separating (EBS) algorithm for extremely fast separation of mixed speech signals without loss of quality, which is performed in two stages: iterative-form separation and closed-form separation. This algorithm significantly improves the separation speed simply due to incorporating only some specific frequency bins into computations. Simulation results show that, on average, the proposed algorithm is 43 times faster than the independent component analysis (ICA) for speech signals, while preserving the separation quality. Also, it outperforms the fast independent component analysis (FastICA), the joint approximate diagonalization of eigenmatrices (JADE), and the second-order blind identification (SOBI) algorithm in terms of separation quality.

• Proceedings of the Computational Structural Engineering Institute Conference
• /
• 2001.10a
• /
• pp.101-108
• /
• 2001

In the previous study, an improved QRD (QR Decomposition)-ILS(Iterative Least-Squares) method is proposed to estimate the structural parameters at the element level using response data alone without using any information of excitation measurements for the assessment of local damages and deterioration in complex and large structural systems. But for a complex and large structural system, where response measurement at every dynamic degree of freedom(DDOF) is not possible, the absence of some observation points of responses and its effect on the proposed SI method must be studied In the paper, a QRD-ILS technique that utilizes the known intact stiffness information estimated based on the visual inspection, field measurements and/or NDT tests is proposed to identify local damages of fracture critical members using measured responses only at limited DDOFs. A numerical example is used to illustrate the application of this technique. The results indicate that the proposed SI technique is very simple but efficient, since no input information are required with only limited observations.