Fast kernel MUSIC algorithm

The rapid progress in miniaturization and integration of micro-structured devices such as integrated circuits has recently increased the demand for a high resolution diagnosis tool. To meet this need, we have developed a highly resolvable fault locator named 3D Laser Microvision, which can nondestructively see through objects with high accuracy. This system is essentially a laser radar that works in a frequency domain rather than a time domain. Instead of sending a narrow pulse into an object, the wavelength of the laser is step-wisely changed ranging from 1.5 /spl mu/m to 1.6 /spl mu/m. At each step, the amplitude and phase of the reflected light from the object are measured and stored in a computer. The inner structure of the object, that is, where refractive indices vary abruptly, is reconstructed by processing these stored data. When a fast Fourier transform is used as the signal processing method, the spatial resolution is 12 /spl mu/m which is limited by the wavelength scanning range. We propose an improved adaptive spatial smoothing technique in the multiple signal classification (MUSIC) algorithm to overcome this limitation and verify its superior performance compared with the conventional MUSIC algorithm by numerical simulation.

We note that the cost function of the MUSIC (multiple signal classification) algorithm is quadratic in an array vector and that it can be expressed in a least squares form. Based on this observation, we present a rigorous Levenberg-Marquardt (LM) formulation of the MUSIC algorithm for simultaneous estimation of an azimuth and an elevation. We show a convergence property and compare the performance of the LM-based MUSIC algorithm with that of the standard MUSIC algorithm via Monte-Carlo simulation. We also compare the performance of the MUSIC algorithm with that of the Capon algorithm both for the standard implementation and for the LM-based implementation.

This study presents a subspace-based tracking algorithm called Bingham filter using the outputs of a sensor array. Given n sensors in the sensor array with spatially and temporally white Gaussian noise in the outputs of the array, the Bingham filter estimates the signal subspaces generated by the targets based on the phased array observations, a priori information and previous estimates. Assuming the complex Bingham distribution as the probability distribution of the target subspace, the authors propose a closed-form solution to the problem of updating the predicted subspace using the phased array observations. Furthermore, the authors offer a closed-form solution for the prediction step of the tracking algorithm based on the relation between the complex Bingham distribution parameters and its covariance matrix. Combining these two steps, a new subspace tracking algorithm is derived that is similar to the Kalman filter. The algorithm is applied to track underwater targets and its efficiency is investigated using simulated phased array observations. Simulation results show that compared with the multiple signal classification (MUSIC) algorithm alone, the combination of the Bingham filter and MUSIC algorithm has better performance in low SNR scenarios.

In this paper, we formulate and experimentally demonstrate direction of arrival (DoA) estimation with a two-port composite right/left handed (CRLH) leaky-wave antenna (LWA) using the multiple signal classification (MUSIC) algorithm. The CRLH LWA consists of a cascade of metamaterial unit cells, periodically modulated using varactor diodes. By changing the voltages across series and shunt varactors, the antenna is able to uniformly steer its radiation pattern. The proposed modified MUSIC algorithm, uses both antenna's input ports to estimate the DoA. Conventionally, the MUSIC algorithm needs to calculate the correlation matrix from the output vector of an array of antenna elements. In this work, we show in detail how the correlation matrix can be computed when using leaky-wave antennas. The algorithm is demonstrated using LWA in an anechoic chamber and the estimated DoA results are in good agreement with the predicted angles.

§This paper presents results obtained as part of a “proof of concept” program being carried out at the University of Nevada, Reno and the NASA Glenn Research Center. The purpose of this program is to determine the feasibility of using passive acoustic methods to determine the performance of jet engines and/or turbines. Data were recorded from six (6) acoustic transducers (microphones) mounted on a combustor ring and installed on a High Pressure Burner Rig (HPBR) located at the NASA Glenn Research Center. The HPBR was operated over several hours using a prescribed “duty cycle” of different fuel/air mixtures. Time frequency signal processing of the data showed that passive acoustic methods could be used to determine fuel/air ratios. These results have been reported in a previous paper. This paper presents additional results obtained by applying other signal processing techniques to the same data set. Correlation techniques and a multiple signal classification (MUSIC) algorithm were applied to the data from each of the microphones. The MUSIC algorithm provided pseudospectrum estimates of the signals by performing eigenspace analysis of the signal’s correlation matrix. The MUSIC algorithm was particularly suited to this data since it is narrow band and has dominant frequencies. Preliminary results presented in this paper indicate that this approach may lead to the ability to generate isothermal maps of a crosssection of a combustor.

To detect inter-harmonics with low amplitude near to harmonics with high amplitude under white noise condition with short-duration data, a method based on interpolation FFT algorithm and Multiple Signal Classification (MUSIC) algorithm is presented in the paper. Firstly, the primary frequency components of signal are extracted from sampled signal by interpolation FFT Algorithm with Hanning window. Residual signal is derived by subtracting the main frequency components of signal from sampled signal. Secondly, the frequency components of residual signal are detected by MUSIC Algorithm. The primary frequency in the first step and the residual frequency in the second step are considered as the whole frequency components of sampled signal. Lastly, the corresponding amplitude and angle of each frequency is detected by least square method. Simulation result show that the proposed method can detect inter-harmonics near the integer harmonics under noise with short data and the estimated result for parameters of frequency and amplitude can meet the need of measuring accuracy.

From the various results of numerical simulations, it has verified that the MUltiple Signal Classification (MUSIC) algorithm can be applied to the limited-view inverse problem. However, this application is somehow heuristic; therefore, the theoretical background of the MUSIC algorithm must be considered beforehand. In this paper, we explore the least condition of MUSIC for guaranteeing the good imaging performance of extended cracks in two-dimensional space. Concerned explore is rely on the expression of imaging function with an infinite series of Bessel functions. Numerical simulations with noisy data verify out investigation.

Joint estimation of azimuth and elevation angles is of great importance in source localization and channel characterization. Firstly, some basic knowledge of three typical parametric estimation algorithms are introduced in this paper, i.e., multiple signal classification (MUSIC), Unitary estimation of signal parameter via rotational invariance technique (ESPRIT), and space-alternating generalized expectation-maximization (SAGE) algorithms. Each algorithm is capable of extracting both azimuth angle of arrival (AAoA) and elevation angle of arrival (EAoA) of multi-paths jointly. It is pointed out that the SAGE and MUSIC algorithms have higher complexity than the Unitary ESPRIT algorithm due to the iteration/angle searching procedure. Secondly, impacts of antenna number, closely spaced paths, and signal-to-noise ratio (SNR) on estimation performance of three algorithms are analyzed. Results show that the Unitary ESPRIT algorithm has lower accuracy in comparison with the MUSIC and SAGE algorithms when antenna number and SNR are large. Finally, three algorithms are applied to estimate multipath parameters in 16 GHz massive MIMO channel measurements. It is shown that the Unitary ESPRIT algorithm performs less satisfactory in MPCs extraction, while MUSIC can provide comparable results with the SAGE algorithm.

Detecting direction of arrival (DoA) using array antenna is considered as common challenge in different typeS of applications, such as sonar, radar, wireless communications. DoA estimation plays an important role in direction finding applications. Further, effective DoA estimation technique is the primary requirements for smart antenna system to realize beam forming technique. Multiple signal classification (MUSIC) and Estimation of Signal Parameters via Rotational Invariance (ESPRIT) algorithms have been widely studied, and they are considered as fundamental for the majority of the DoA methods in literature. In this paper, the principal of DoA estimations is presented. The main focus of the current paper is on the classical methods (Bartletts and MVDR), MUSIC and ESPRIT algorithms. Review of various efforts that have been made to enhance the performance of these algorithms was acheived. Then compared their performances using simulation results. Further, applying MUSIC algorithm to estimate the frequency of the multiple incident signals. The results show that MUSIC algorithm has higher resolution compared to other presented techniques. It is also shown that the accuracy of ESPRIT method is mainly rely on the number of antenna elements. Finally, the principle of Eigenvalues-based method was proposed. It is shown that this method has good resolution with very low complexity as it does not need decomposition of covariance matrix.

In this paper, we address the problem of tracking the direction (DOA) of coherently distributed (CD) sources using subspace updating. The conventional multiple signal classification (MUSIC) algorithm, which should estimate covariance matrix, can not perform DOA tracking until it acquires covariance matrix. In addition, MUSIC algorithm needs spectrum peaks search and inverse matrix operator, so it can not estimate DOA of distributed sources rapidly. By contrast, the proposed algorithm uses subspace tracking to estimate signal subspace iteratively, then, the central angles are estimated using TLS-ESPRIT for CD sources. This algorithm can rapidly track DOA of CD sources even when their angular spreads are slightly wide

Two techniques are proposed to alleviate the computational burden of MUltiple SIgnal Classification (MUSIC) algorithm applied to Electroencephalogram (EEG) source localization. A significant reduction was achieved by parsing the cortex surface into smaller regions and nominating only a few regions for the exhaustive search inherent in the MUSIC algorithm. The nomination procedure involves a dictionary learning phase in which each region is assigned an atom matrix. Moreover, a dimensionality reduction step provided by excluding some of the electrodes is designed such that the Cramer-Rao bound of localization is maintained. It is shown by simulation that computational complexity of the MUSIC-based localization can be reduced up to 80%.

This paper investigates the passive radar imaging system which exploits digital video broadcasting (DVB) satellites as opportunity illuminators. Firstly, we establish the imaging model and then analyze the imaging performance in the wavenumber domain. Because of the practical limitations, we are likely to get an undersampled wavenumber domain coverage so that traditional imaging methods such as matched filter (MF) would fail to provide good imaging performance. Therefore, we consider to use the sparse recovery techniques to achieve good inversion results. However, the compressive sensing (CS) imaging methods would be severely affected when off-grid scatterers exist. According to the theories of the Xampling and the finite rate of innovation (FRI), we propose a high-resolution sparse imaging method to overcome the problems above, which is based on the multiple signal classification (MUSIC) algorithm. Simulation results verify the effectiveness of the proposed method.

Blade tip timing (BTT) is a core technique for investigating the blade vibration of large-scale centrifugal compressors, and identifying the parameters of blade asynchronous vibration is crucial for implementing blade condition monitoring based on the BTT technique. In this study, the multiple signal classification (MUSIC) algorithm and the estimating signal parameters via rotational invariance techniques (ESPRIT) algorithm were first applied separately to identify the asynchronous vibration parameters of centrifugal compressor blades, with their advantages and disadvantages discussed. Subsequently, based on the frequency distribution characteristics in the ESPRIT results, the concept of “frequency distribution rate” was proposed. Finally, the results of the MUSIC algorithm were weighted by the frequency distribution rate, and an improved MUSIC algorithm was proposed. This enhanced confidence in the real frequency in the MUSIC algorithm results. Compared with the strain gauge method, the maximum relative error of the improved algorithm is 0.23%. The improved MUSIC algorithm improves the accuracy of parameter identification for blade asynchronous vibration, which holds great significance for the industrial application of the BTT technique.

A focusing MUSIC algorithm for baseline-free Lamb wave damage localization

This paper presents a brief overview of narrowband direction of arrival (DOA) estimation algorithms and techniques. A comprehensive study is carried out in this paper to investigate and evaluate the performance of variety of algorithms for DOA estimation. Two categories of DOA estimation algorithms are considered for discussion which are Classical methods and Subspace based techniques. Classical methods include Sum-and-Delay method and Capon’s Minimum Variance Distortionless Response (MVDR) while Subspace based techniques are multiple signal classification (MUSIC) and The Minimum Norm Technique. Also ESPIRIT technique is evaluated. Inefficiencies are pointed out and solutions are suggested to overcome these shortfalls. Simulation results shows that the MUSIC algorithm is able to better represent the DOAs of signals with more prominent peaks. The Min-Norm algorithm also identifies the DOAs of signals similar to the MUSIC algorithm, but produces spurious peaks at other locations. The MVDR method identifies the DOAs of signals, but the locations are not represented by sharp peaks, due to spectral leakage. The classical beamformer also produces several spurious peaks. MUSIC show higher accuracy and resolution than the other algorithms. It should be noted that MUSIC is more applicable because it can be used for different array geometries.