Direction-of-arrival Estimation of Wideband Sources Using Sensor Arrays

윤여선 Georgia Institute of Technology 2004 해외박사

Sensor arrays are used in many applications where their ability to localize signal sources is essential. For many applications, it is necessary to estimate the direction-of-arrival (DOA) of target sources. Although there are many DOA estimation methods available, most of them are valid only for narrowband signals where time delay can be approximated as a phase shift. This thesis focuses on DOA estimation algorithms for wideband sources. Specifically, this thesis proposes the pruned fast beamformer which can reduce the number of computations of delay-and-Sum (DS) beamforming by using a multi-resolution structure. For high resolution methods, signal subspace methods are required. Most of the subspace techniques for wideband signals decompose the received wideband signals into several bands of narrowband signals through bandpass filtering. Then, there are two different ways of processing decomposed signals. The incoherent methods process each band independently by a given narrowband method and average the results. The coherent methods attempt to modulate the signals in each band so that they can be combined coherently. In this thesis, a new DOA estimator, which is called TOPS, is developed to avoid disadvantages of both the incoherent and the coherent methods. The new method which can be categorized as a non-coherent method is tested and compared with other methods. It exhibits many desirable features for a number of applications where the sources are wideband such as acoustic direction finding.

Deep Learning-Aided Coherent Direction-of-Arrival Estimation with The FTMR Algorithm

호안 뜨렁 다이 서울과학기술대학교 2021 국내석사

In this work, we apply deep learning to estimate the direction-of-arrival (DoA) of multiple narrowband signals with a uniform linear array in a coherent environment. First, the logarithmic eigenvalue-based classification network (LogECNet) is introduced to enhance the signal number detection accuracy. Next, a multi-label classification model called the root-spectrum network (RSNet) is devised to estimate the DoAs using the signal number inferred by LogECNet. In the proposed architecture, the full-row Toeplitz matrices reconstruction (FTMR), which exploits all rows of the signal covariance matrix (SCM), is combined with LogECNet and RSNet to inversely map the SCM to the numerical DoAs in the coherent scenario. It is shown that the eigenvalues factorized from the FTMR output matrix become more robust sources for signal enumeration than those of the forward/backward spatial smoothing (FBSS) algorithm for signal enumeration. Furthermore, the logarithmic scaling of the eigenvalues of the FTMR results in LogECNet outperforming other detectors. The simulation results show our proposed method not only improves the signal number detection and angular estimation performance, but also achieves the complexity reduction with respect to the prior schemes. 본 논문에서는 균일선형배열(Uniform Linear Array)을 이용한 코히런트(Coherent) 다중 협대역 신호의 도래각(Direction-of-Arrival) 추정에 심층학습(Deep Learning)을 적용한다. 이를 위해 첫번째로 신호 개수 추정 정확도를 높이는 LogECNet(Logarithmic Eigenvalue-based Classification Network)을 제안한다. 두번째로 LogECNet에서 추론한 신호 개수를 사용하여 도래각을 추정하는 다중 레이블 분류 모델인 RSNet(Root-Spectrum Network)을 제안한다. 제안된 아키텍처는 신호 공분산 행렬의 전체 행을 활용하는 FTMR(Full-row Toeplitz Matrices Reconstruction)을 LogECNet 및 RSNet과 결합하여 코히런트 환경에서 신호 공분산 행렬을 도래각에 매핑한다. FTMR 출력 행렬에서 얻어지는 고유값(Eigenvalue)들은 신호 개수 검출에서 FBSS(Forward/Backward Spatial Smoothing) 알고리즘보다 더 강인한 입력원으로 동작한다. 또한 FTMR 행렬의 고유값을 로그함수를 통해 스케일링(Scaling)함으로써 LogECNet이 기존 검출기보다 높은 성능을 얻도록 한다. 모의 실험 결과를 통해 제안한 방식이 신호 개수 검출 및 도래각 추정 성능을 향상시킬 뿐 아니라 이전 방식에 비해 낮은 복잡도를 요구함을 확인하였다.

Direction-of-Arrival estimation of multiple sound sources in noisy and reverberant environments using Laplacian Mixture Model : 반향이있는잡음환경에서라플라스혼합모델을사용한다중음원의방향추정

누엔황안 배재대학교 일반대학원 2011 국내석사

Direction-of-Arrival (DOA) estimation of multiple sound sources or sound source localization is a basic but important technique to many applications such as robots, speech enhancement, and surveillance. Many efficient methods for DOA estimation have been studied so far, which include Generalized Cross Correlation (GCC) based method, time-frequency histogram, Degenerate Unmixing Estimation Technique (DUET), Kernel Density Estimation (KDE) and so on. In this thesis, we investigate a probabilistic method to estimate the DOA of each source in speech mixtures in noisy reverberant environments. Laplacian mixture model(LMM) is used to model the speech mixture and log-likelihood function is formulatedto find the DOA of each source in a mixture. Since the log-likelihoodfunction is a function of hidden variables, we cannot obtain the solution directly. Instead, an Expectation-Maximization (EM) algorithm is implemented to find thesolution iteratively. In the simulation section, our method was implemented in both batch and online processing modes. We used a two-microphone array with inter-microphone distance of 15㎝. To mimic noisy reverberant environments, speech samples from TIMIT data base were convolved with impulse responses generated by RoomSim MATLAB toolbox and noise were added to get 10㏈ signal-to-noise ratio. Each mixture consists of 2~4 sources that are heavily overlapped. In online processing mode, both fixedand moving sound sources were tested. We compared the performance of LMM and Gaussian mixture model (GMM). We found that the LMM is more accurate in DOA estimation of multiple sources than the GMM in all cases simulated. 다중음원의 방향 추정(DOA) 혹은 음원 위치 확인은 로봇, 음질 개선, 환경 감시 등의 응용에 기본적인 중요한 요소이다. 현재까지 Generalized Cross Correlation (GCC), 시간-주파수 히스토그램, Degenerate Unmixing Estimation Technique (DUET), Kernel Density Estimation (KDE)등 같은 여러 방식이 연구 제안되었다. 그러나 이들의 성능은 잡음과 반향의 정도에 따라 성능이 저하되는 단점이 있다. 이 논문에서는 잡음이 많고 반향이 많은 환경에서 효과적인 다중 음원의 DOA 추정 방법을 연구한다.마이크로폰에 수집된 음성 혼합물는라플라스혼합모델로 모델되고 이로부터 각 음원의 DOA를 추정하기 위해 로그-우도함수를 수식화한다.로그-우도함수는 알려지지 않은 변수의 함수이므로 직접적으로 해를 구할 수 없다. 대신, Expectation-Maximization (EM) 알고리즘을 이용하여 반복적인 방법으로 해를 구한다. 방법의 효율성을 검증하기 위해 배치 온라인 처리 과정이 모두 구현되었다. 마이크로폰 사이의 거리가 15㎝인 두 개의 마이크로폰으로 구성된 어레이를 사용하였다. 반향이 있는 잡음환경을 만들기 위해RoomSim MATLAB toolbox로부터 임펄스 응답을 만들었으며, 이를 TIMIT 데이터베이스의 음성 신호 샘플들과 콘볼루션한 다음 SNR이 10 ㏈가 되도록 잡음을 추가하여 마이크로폰 신호를 만들어 냈다. 각각의 혼합물은 2~4개의 중첩도가 매우 높은 음성으로 구성되어 있다. 온라인 모드에서는 음원이 고정된 경우와 움직이는 경우를 모두 고려하였다. 실험을 통해 라플라스 모델과 가우시안 모델의 성능을 비교하였으며, 거의 모든 경우에서 라플라스 모델이 더 정확하고 계산이 간단함을 확인할 수 있었다.

(A) study on the compressive sensing based high resolution methods and AOA-based three-dimensional localization

백지웅 세종대학교 대학원 2021 국내박사

The compressive sensing based direction-of-arrival estimation algorithm is a high-resolution method that estimates the incident angles of multiple targets by using the spatial sparsity of the incident signal. The compressive sensing based direction-of-arrival estimation method has the advantage of being more robust to the phased array failure and noise than beamforming-based methods and subspace-based methods with high resolution. In the case of a localization method that estimates the location coordinates of a threaten target based on the angle-of-arrival, it is very important research topic for collecting location information of threats that do not provide GPS information. The compressed sensing-based super-resolution methods have been studied with a focus on the passive phased array system and the bistatic MIMO system. The first proposed scheme is an optimal weight regularization parameter selection method of a covariance fitting algorithm with respect to the number of snapshots. A covariance fitting algorithm for the estimation of direction-of-arrivals (DOAs) of multiple incident signals is addressed in this dissertation. The scheme takes advantage of the fact that the incident signals are spatially sparse. A previous study has presented the regularization parameters of the covariance fitting for a very large number of snapshots. In this dissertation, a strategy on how to determine the regularization constant of the covariance fitting for a general number of snapshots is presented. The strategy essentially exploits the norm of the noise covariance matrix. The proposed algorithm has been validated via numerical simulations. The second proposed scheme is an enhanced smoothed l0-norm algorithm for the passive phased array system, which uses covariance matrix of the received signal. The SL0 (smoothed l0-norm) algorithm is a fast compressive sensing-based DOA(direction-of-arrival) estimation algorithm that uses a single snapshot from the received signal. In the conventional SL0 algorithm, there are limitations in the resolution and the DOA estimation performance, since a single sample is used. If multiple snapshots are used, the conventional SL0 algorithm can improve performance in terms of the DOA estimation. In this dissertation, a covariance fitting based SL0 algorithm is proposed to further reduce the number of optimization variables when using multiple snapshots of the received signal. A cost function and a new null-space projection term of the sparse recovery for the proposed scheme are presented. In order to verify the performance of the proposed algorithm, we present the simulation results and the experimental results based on the measured data. The third proposed scheme is a compressive sensing-based data fitting direction-of-departure/direction-of-arrival (DOD/DOA) estimation algorithm which is proposed to apply the superior performance of the compressive sensing method to the bistatic MIMO systems. The algorithm proposed in this dissertation optimizes the output data via convex optimization-based sparse recovery, so that it is possible to estimate the DOD and the DOA for each target accurately. In order to minimize the amount of computation, the cost function with constraint condition is implemented in this paper. Furthermore, the constraint condition parameter of the cost function is analytically derived. Through various simulations, it is shown that the superior DOD and DOA estimation performance of the proposed algorithm and that the analytical derivation of the constraint condition parameter is useful for determination of regularization parameter. The fourth proposed scheme is a three-dimensional linear least square based localization algorithm. Closed-form expression of three-dimensional emitter location estimation using azimuth and elevation measurements at multiple locations is presented in this dissertation. The three dimensional location estimate is obtained from three dimensional sensor locations and the azimuth and elevation measurements at each sensor location. Since the formulation is not iterative, it is not computationally intensive and does not need an initial location estimate. Numerical results are presented to show the validity of the proposed scheme.

Efficient parameter estimation methods for automotive radar systems

이한별 서울대학교 대학원 2016 국내박사

As the demand for safety and convenience in the automotive-technology field increased, many applications of advanced driving assistance systems were developed. To provide driving information, among the sensors, such as cameras sensor, light detection and ranging sensor, radar sensor, and ultrasonic sensor, a radar sensor is known to exhibit excellent performance in terms of visibility for different weather conditions. Especially with the legislation of the adaptive cruise control system and autonomous emergency braking system in a global environment, the market of the automotive radar sensor is expected to grow explosively. At present, the development of cost-effective radar offering high performance with small size is required. In addition, the radar system should be enforced to have a simultaneous functionality for both long and short ranges. Thus, challenging issues still remain with respect to radar signal processing including high-resolution parameter estimation, multi-target detection, clutter suppression, and interference mitigation. For high-resolution parameter estimation, direction-of-arrival (DOA) estimation method has been investigated to identify the target object under complex unban environment. To separate closely spaced target having similar range and distance, high-resolution techniques, such as multiple signal classification (MUSIC), the estimation of signal parameters via rotational invariance techniques (ESPRIT), and maximum likelihood (ML) algorithm, are applied for automotive radars. In general, cycle time for radar system, which is the processing time for one snapshot, is very short, thus to establish a high-resolution estimation algorithm with computational efficiency is additional issue. On the other hands, multi-target detection scheme is required to identify many targets in the field of view. Multi-target detection is regarded as target pairing solution, whose task is to associate frequency components obtained from multiple targets. Under certain conditions, the association may fail and real target may be combined to ghost components. Thus, reliable paring or association method is essential for automotive radar systems. The clutter denotes undesired echoes due to reflected wave from background environment, which includes guardrail, traffic signs, and stationary structures around the load. To minimize the effect of clutter, conventional radar systems use high pass filter based on the assumption that the clutter is stationary with energy concentrated in the low frequency domain. However, the clutter is presented with various energy and frequency under automotive radar environment. Especially, under the specific environment with iron materials, target component is not detected due to clutter with large power. Mutual interference is a crucial issue that must be resolved for improved safety functions. Given the increasing number of automotive radar sensors operating at the same instant, the probability that radar sensors may receive signals from other radar sensors gradually increases. In such a situation, the system may fail to detect the correct target given the serious interference. Effective countermeasures, therefore, have to be considered. In this dissertation, we propose efficient parameter estimation methods for automotive radar system. The proposed methods include the radar signal processing issues as above described, respectively. First, the high-resolution DOA estimation method is proposed by using frequency domain analysis. The scheme is based on the MUSIC algorithm, which use distinct beat frequency of the target. The target beat frequency also gives distance and velocity. Thus, the proposed algorithm provides either high-resolution angle information of target or natural target pairing solution. Secondly, we propose the clutter suppression method under iron-tunnel conditions. The clutter in iron-tunnel environments is known to severely degrade the target detection performance because of the signal reflection from iron structures. The suppression scheme is based on cepstral analysis of received signal. By using periodical characteristic of the iron-tunnel clutter, the suppressed frequency response is obtained. Finally, the interference mitigation scheme is studied. Mutual interference between frequency modulated continuous waveform (FMCW) radars appears in the form of increased noise levels in the frequency domain and results in a failure to separate the target object from interferer. Thus, we propose a high-resolution frequency estimation technique for use in interference environments.

AI-BASED RADAR SIGNAL PROCESSING

고찬빈 세종대학교 대학원 2020 국내석사

In this paper, machine learning based radar signal processing is addressed. First, A weather radar signal detection was conducted. Due to the problem of exhaustion of frequency resources, many studies on frequency sharing have been conducted. In order to share the frequency of WLAN(Wireless LAN) and weather radar, weather radar signal detection is essential. The performance of the weather radar signal detection was improved using machine learning based algorithms such as SVM(Support Vector Machine) and CNN(Convolutional Neural Network). Secondly, the DOA(direction-of-arrival) estimation of FMCW(Frequency Modulated Continuous Wave) radar was conducted. When the number of sensors is limited, conventional algorithms such as MUSIC(Multiple Signal Classification) have low angle resolution for performing DOA estimation. It is important to distinguish closely spaced targets on the road. In this paper, the angle resoulution was improved using DNN(Deep neural network)- based DOA estimation.

A Study on Frequency Domain Implementation and Performance of Compressive Sensing-Based Algorithm

ZHANG XUEYANG 세종대학교 대학원 2018 국내석사

Direction-of-arrival (DOA) estimation is widely used for communication, radar signal processing, sonar signal processing and etc. DOA estimation is a technique for estimating the DOA of the signal source using the signal data received from the sensor array, and it is an important factor of locating the position of the signal source. In this paper, the Compressive Sensing-based(CS) algorithm is studied, which is a DOA estimation algorithm. In chapter 1, the sensor array structure that used in the DOA estimation to receive the signal data is described. And the typical DOA estimation algorithms, like Conventional Beamforming (CBF), Minimum Variance Distortionless Response (MVDR), and Multiple Signal Classification (MUSIC)), are studied. In chapter 2, the implementation of the CS-based algorithm in the frequency domain was described. The published papers on DOA estimation using the CS-based algorithm have been dealt with in the time domain. In order to extend the CS-based algorithm into the frequency domain, the received signal data is transformed to the frequency domain through Fast Fourier Transform (FFT). And using the characteristics of the tone signal, the signal data in the frequency domain can be obtained. By applying the data to the DOA estimation algorithm, the DOA estimation of the incident signal in the frequency domain is implemented. In chapter 2, the data fitting algorithm and the covariance fitting algorithm were described, which are typical CS-based DOA estimation algorithms. In order to verify the DOA estimation performance in the frequency domain, the simulation was performed by using CBF, MVDR and the CS-based algorithm, and the DOA estimation performance of each algorithm was shown by calculating the mean square error(MSE) of each algorithm. It is shown that the CS-based algorithm has superior DOA estimation performance than other DOA estimation algorithms. In general, the sensor elements are arranged at a maximum half-wave interval. In the actual underwater environment, because the geographical factors, the sensor elements are arranged at a non-uniformly interval, or some sensors in a uniform linear array fails and that results in a uniform sensor array becoming a non-uniform sensor array. And the non-uniform sensor array is used to received signal data. In chapter 3, the DOA estimation by using data fitting algorithm in those special sensor array environment was described. 도래각 추정은 통신, 레이더 신호 처리, 소나 신호 처리 등에 널리 사용되고 있다. 도래각 추정은 센서 배열로부터 수신한 신호 데이터를 사용하여 신호원의 도래각을 추정하는 기법이고 신호원 위치를 추정하는 데 있어서 중요한 요소이다. 본 논문에서는 압축센싱 기반 도래각 추정 알고리즘을 다룬다. 본 논문의1 장에서는 도래각 추정 시 이용되는 센서 배열 구조 와 대표적인 도래각 추정 알고리즘 (Conventional Beamforming (CBF), Minimum Variance Distortionless Response (MVDR) 그리고 Multiple Signal Classification (MUSIC))을 다루었다. 2 장에서는 주파수 영역으로 확장된 압축센싱 기반 알고리즘에 대해 연구한다. 기존에 발표된 압축센싱 기반 도래각 추정 알고리즘에 관한 논문들은 시간 영역에 대해 다루었다. 압축센싱 기반 도래각 추정 알고리즘을 주파수 영역으로 확장하기 위해, 센서 배열로부터 수신 된 톤 신호 데이터는 고속 푸리에 변환 (FFT)하여 주파수 도메인으로 시킨다. 톤 신호의 특성을 사용함으로써, 주파수 영역의 수신 데이터를 얻는다. 그리고 해당 데이터를 도래각 추정 알고리즘에 적용함으로써 신호의 도래각을 추정한다. 2장에서는 압축센싱 기반 도래각 추정 알고리즘인 데이터 fitting 알고리즘과 공분산 fitting 알고리즘에 대해 다루었다. 주파수 영역으로 확장한 압축센싱 도래각 추정 알고리즘의 추정 성능을 확인하기 위해 CBF, MVDR 그리고 압축센싱 기반 도래각 추정 알고리즘을 사용하여 시뮬레이션을 수행하고 각 알고리즘의 추정 성능을 보인다. Mean-square Error(MSE)을 통해 SNR에 따른 각 알고리즘의 도래각 추정 성능을 비교했다. 시뮬레이션을 통해 구현한 압축센싱 기반 도래각 추정 알고리즘이 다른 방위각 추정 알고리즘에 비해 우수한 방위각 추정 성능을 가지고 있을 보였다. 3 장에서는 특수한 경우의 센서 배열 환경에서의 데이터 fitting 도래각 추정 알고리즘의 추정 성능에 대해 다루었다. 센서 배열의 구조는 도래각 추정 성능에 대한 영향을 준다. 일반적으로 센서 소자들은 최대 반 파장 간격으로 배치된다. 하지만 실제 수중 환경 에서는 지형적인 요소 등으로 인해 센서 소자가 비 균일 하게 설치되어 신호의 정보를 수신하는 경우가 있다. 또한, 선형 배열 중의 일부 센서가 고장하여 균일 한 선형 배열이 비 균일 배열이 되는 경우가 있다. 3 장에서는 위의 두 가지 경우 환경에서 데이터 fitting 알고리즘을 구현 하고 이를 이용하여 도래각을 추정을 수행하고 이에 대한 성능을 보인다.

Towards autonomous driving : surrounding environment classification and DOA estimation via advanced radar signal processing

심헌교 서울대학교 대학원 2020 국내박사

최근 들어, 자율 주행 자동차와 관련된 관심이 높아지면서 자율 주행에 사용되는 센서들과 관련된 연구가 활발히 진행되고 있다. 자율 주행에 사용되는 센서에는 카메라, 라이다, 레이더, 초음파 등이 있는데, 그 중에서 레이더는 최대 탐지 가능 거리가 길고, 빛이 없는 상황이나 비가 오는 상황 등 열악한 환경에 강인한 특성을 가지고 있어서 자율주행에 필수적이다. 레이더는 다양한 용도로 사용이 가능한데, 주로 적응형 순항 제어와 자동 긴급 제동에 사용된다. 레이더는 원하는 타깃까지의 거리, 상대속도, 각도 등을 탐지할 수 있고, 타깃의 종류, 크기 등에 대해서도 탐지가 가능하다. 본 학위 논문에서는 도로 환경을 분류할 수 있는 인공 신경망 구조를 제안하였다. 자율주행을 하다 보면 다양한 도로 환경을 마주하게 되는데, 도로 환경에 따라 그에 맞는 타깃 검출 알고리즘을 적용할 필요가 있다. 예를 들어, 여러 개의 철제 구조물로 이루어진 철제 터널의 경우에는 반사 신호가 아주 강해서 타깃이 묻히게 되는 현상이 발생하게 된다. 따라서, 원하는 타겟을 검출하기 위해서는 클러터 제거 알고리즘을 적용하는 것이 필요하다. 본 논문에서는 각각의 도로 환경에 적합한 알고리즘을 적용하기 위한 사전 작업으로 딥러닝 기법을 이용하여 도로 환경을 분류하고, 도로 환경을 멀리서도 미리 인식하는 인공 신경망의 구조 제안하였다. 결과적으로, 도로 환경 분류 정확도를 약 14%p 정도 향상시켰다. 또한, 차량용 레이더를 이용하여 각도 분해능을 향상 시킬 수 있는 기법을 제안하였다. 타깃의 위치 정보를 알기 위해서는 거리, 속도 정보 이외에도 각도 정보가 필수적이다. 하지만, 서로 다른 두 타깃이 가까이에 위치하게 되면, 각도 분해능의 한계로 인해 두 타깃이 한 타깃으로 나타나는 현상이 발생하게 된다. 이는 안테나 개구면의 크기를 늘려서 해결이 가능하지만 개구면의 크기를 늘리게 되면 시야각이 줄어들고, 물리적인 공간도 많이 차지하게 되는 단점이 있다. 이를 해결하기 위해 본 논문에서는 선형 예측 안테나 외삽 방식을 이용하여 가상의 수신 신호를 생성하는 방식을 제안하였다. 생성된 가상의 수신 신호와 실제 신호를 도래각 추정 알고리즘에 적용하여 작은 안테나 개구면의 크기를 사용하면서도 각도 분해능은 향상시키는 방법을 제안하였다. 결과적으로, 제안한 기법을 이용하여 각도 분해능을 약 3˚ 가량 향상시켰다. 마지막으로, 레이더에서의 송신 신호 분류 기법을 제안하여 타깃의 위치 추정 성능을 향상시켰다. 여러 개의 송신 안테나를 사용하게 되면 적은 개수의 안테나를 이용하여 타깃을 효율적으로 검출할 수 있다. 하지만, 각각의 송신 안테나로부터 방사된 신호를 구분하지 못하게 되면, 각도 추정 성능에 열화가 발생한다. 이러한 문제를 해결하기 위해 최대 우도 추정 기법을 이용하여 송신 신호를 분류하는 기법을 제안하였다. 결과적으로, 최대로 탐지 가능한 속도 범위를 2배로 증가시켰다. 또한, 송신 신호를 구분하는 기법을 이용하여 각도의 평균 제곱은 오차를 3˚ 정도 향상시켰다. Recently, as interest in autonomous driving has increased, research on sensors used for autonomous driving is being actively conducted. These sensors used include cameras, lidars, radars, and ultrasonic devices. Among them, radars have a long maximum detectable range and are robust to harsh environments such as rain or no light. Thus, they are essential sensors for autonomous driving. Radars can be used for various purposes, mainly for adaptive cruise control (ACC) and automatic emergency braking (AEB). They can be used to detect the distance to the target, relative velocity, angle, etc., and they can also detect target type and size. In this dissertation, I proposed a neural network structure for classifying road environments. Various road environments are encountered in autonomous driving, and applying an appropriate target detection algorithm depending on the road environment is necessary. For example, in an iron tunnel comprising several iron structures, the reflection signal is very strong, causing the target to be undetected. Therefore, to detect a desired target, a clutter removal algorithm should be required. To recognize and classify the road environment from a distance in advance, I proposed a neural network structure. As a result, the accuracy of classifying road environments was improved by approximately 14%p. In addition, I proposed a method to improve the angular resolution by using an automotive radar. To know the position of the target, the direction-of-arrival (DOA) as well as distance and velocity are essential information. This can be solved by increasing the antenna aperture size; however, increasing the aperture size reduces the field of view and requires considerable space. To solve this problem, I proposed a method for generating virtual received signals using the linearly predicted array expansion. By applying the generated virtual and actual signals to the DOA estimation algorithm, a method of improving the angular resolution using a small antenna aperture size was proposed. As a result, the proposed method improved the angular resolution by approximately 3˚. Finally, I proposed a technique to distinguish the transmission signal in the multi-input and multi-output (MIMO) radar to improve the DOA estimation performance. If multiple transmitting antennas are used, the targets can be efficiently detected by using a small number of antennas. However, if the signals radiated from each transmit antenna element cannot be distinguished, the DOA estimation performance degrades. To address this problem, I proposed a method to distinguish transmission signals by using a maximum likelihood estimation method. Thus, the maximum detectable velocity was doubled. In addition, the DOA estimation performance was also enhanced by approximately 3˚ in terms of root mean square error (RMSE).

Direction finding based on time-modulated circular array for full azimuthal coverage

이동욱 Graduate School, Yonsei University 2021 국내석사

In this thesis, the direction finding (DF) method using the time-modulated circular array (TMCA) with directional elements is proposed to cover the full azimuthal plane. The proposed TMCA DF system has the advantages of simple structure, low complexity on signal processing, and relatively high accuracy compared to amplitude, phase, and spatial spectrumbased DF methods. The proposed system is composed of a uniform circular array with unidirectional antennas and RF switches, and a digital signal processor (DSP) for the direction of arrival (DoA) estimation. In order to estimate the direction of arrival (DoA), harmonic characteristics comparison in the frequency domain and amplitude comparison in the time domain are executed parallelly in DSP. The sector containing the signal is determined by comparing the amplitude of the received signal over time, and the direction is estimated by comparing the harmonic characteristics obtained through the fast Fourier transform (FFT). The phase ambiguity caused by the symmetrical pattern of the circular array can be removed without additional hardware through the proposed DF method. Furthermore, the proposed method is verified by DF simulation and experimental results. The DF simulation results have the root mean square error (RMSE) of about ±1.2° in the range of azimuth plane 360°. According to simulation results, the proposed method has relatively high accuracy compared to other DF methods. For the DF experiment, the 4-element TMCA DF system is fabricated consisting of a microstrip patch antenna, a single-pole fourthrow (SP4T) RF switch, and FPGA to control the switch. The experimental results have a mean deviation error of about ±2.93°. Although the result shows a slightly higher error compared to the simulation result, it could be improved by compensating for the imperfections of the fabricated TMCA DF system. The proposed system reduces the complexity and cost of hardware and has relatively high estimation accuracy since the RF switches with the single RF channel are used instead of the phase shifters or the phase detectors. Therefore, the proposed TMCA DF system could be used in various applications, which is required a low cost, low complexity, and real-time DF system with high accuracy. 본 논문에서는 방위각 360°범위의 임의의 신호에 대한 방향을 추정할 수 있는 time-modulated circular array (TMCA)를 이용한 방향 탐지 시스템을 제안한다. 제안된 TMCA 방향 탐지 시스템은 진폭, 위상, 공간 스펙트럼 방향 탐지 방식에 비해 간단한 구조와 신호 처리의 복잡도가 낮으며, 상대적으로 높은 정확도의 장점을 가진다. TMCA 방향 탐지 시스템의 하드웨어 단은 단방향 지향성 안테나로 구성된 원형 배열과 RF 스위치로 이루어져있으며, 신호 처리 단은 주파수 영역에서의 고조파 특성 비교와 시간 영역에서의 진폭 비교로 구성되어있다. 배열에 입사된 신호는 각 안테나에 연결된 RF 스위치에 의한 순차적으로 스위칭에 의해 time-modulation 된 신호는 단일 RF 채널을 통해 기저 대역으로 전달된다. 수신된 신호의 시간에 따른 진폭 비교를 통해 신호가 있는 영역을 결정하고, 고속 푸리에 변환을 통한 얻은 고조파 특성 비교를 통해 입사된 신호의 방향을 추정한다. 두 가지 방식을 병렬로 처리함으로써 입사된 신호의 각을 도출할 수 있으며, 고조파 특성 분석만을 사용했을 때 발생하는 원형 배열 구조의 대칭되는 패턴으로 인한 위상 모호성을 추가적인 하드웨어없이 제거할 수 있다. 제안된 TMCA 방향 탐지 시스템을 검증하기 위해 방향 탐지 시뮬레이션 및 실험을 진행하였다. 방향 탐지 시뮬레이션 결과에 의해 평균 제곱근 오차 (RMSE)는 방위각 360°범위에서 약 1.2°를 나타냈다. 이를 통해 제안된 방향 탐지 시스템은 다른 방향 탐지 방법에 비해 상대적으로 높은 정확도를 가지는 것을 확인하였다. 방향 탐지 실험을 위해 마이크로스트립 패치 안테나, singlepole four-throw (SP4T) 스위치 및 스위치 제어를 위한 FPGA으로 구성된 4-요소 TMCA 방향 탐지 시스템을 구현하였다. 실험 결과에 의해 제안된 방향 탐지 시스템의 평균 오차는 약 ±2.93°를 나타냈다. 시뮬레이션 결과에 비해 다소 높은 오차를 나타내지만, 구현한 하드웨어의 매칭과 스위치 회로 등 시스템의 불완전성을 보정함으로써 개선이 가능함을 확인하였다. 제안된 방향 탐지 시스템은 위상 천이기 및 위상 검출기를 RF 스위치로 대체하고 단일 RF 채널을 사용함으로써 하드웨어의 비용 및 복잡도가 감소할 뿐 아니라 상대적으로 높은 방향 탐지 정확도를 가진다. 따라서 제안 된 TMCA 방향 찾기 시스템은 저비용, 낮은 복잡성, 높은 정확도의 실시간 방향 탐지가 요구되는 다양한 응용 분야에서 사용이 가능할 것으로 보인다.

(An) autonomous path keeping system based on camera and wireless sensor networks

박종록 Graduate School, Yonsei University 2012 국내박사