Unsupervised Learning of Higher-Order Statistics

This paper deals with the adaptive extraction of higher-order statistics of related signals. We will show how to use higher-order neural networks to adaptively extract the higher-order cumulant matrices and tensors with an invariant weight norm. This proposed scheme can serve as an alternative tool in many application fields with higher-order statistics.     

联合时序结构和负熵的FECG盲提取算法

在提取胎儿心电信号时,利用时序结构的盲源提取算法对胎心率误差估计敏感;利用负熵的盲源提取算法存在提取出的信号具有不确定性问题.为克服这两种算法的缺点,利用时序结构和负熵构造了联合代价函数,利用梯度算法对该代价函数求偏导,提出了一种新的胎儿心电提取算法.该算法能确保每次提取出来的信号都是胎儿心电信号,并且在胎心率估计存在较大误差和胎心率变化的情况下,也能正常工作.计算机仿真和实际胎儿心电提取实验验证了该算法在提取胎儿心电信号时的稳定性和确定性.     

A Spurious Equilibria-free Learning Algorithm for the Blind Separation of Non-zoer Skewness Signals

We present a new learning algorithm for the blind separation of independent source signals having non-zero skewness (the 3rd-order cumulant) (the source signals have non-symmetric probability distribution.), from their linear mixtures. It is shown that for a class of source signals whose probability distribution functions is not symmetric, a simple adaptive learning algorithm using quadratic function (f(x)=x²) is very efficient for blind source separation task. It is proved that all stable equilibria of the proposed learning algorithm are desirable solutions. Extensive computer simulation experiments confirmed the validity of the proposed algorithm.     

一种奇异混合信号盲分离的神经网络模型

混合信号盲分离问题是一类很难而又具有很强应用背景的问题,以往对这类问题的研究均在混合矩阵为非奇异的条件下进行,本文给出一种神经网络模型及相应算法,处理混合矩阵为奇异的混合信号盲分离问题。     

基于约束反赫布算法的自适应二维FIR滤波器实现

提出了一种用神经网络来实现自适应二维FIR滤波器的新方法,主要是运用约束反赫布学习算法,训练一个神经网络,实现了基于TLS准则的自适应二维FIR滤波器,计算机仿真结果表明该滤波器在收敛性能和精度上均优于基于LMS准则的自适应二维FIR滤波器。     

一种基于高阶矩的OFDM信号调制盲识别算法

提出一种Rayleigh衰落信遭务件下的多载波调制盲识别算法,用以区分多载波调制信号（如OFDM）和数字单载波调制信号（如MPSK,MQAM,MFSK）。该算法不需要预先知道信号的载波频率和波特率,只需从中频信号直接进行识别处理。算法中利用信号的高阶统计量作为分类特征,采用信噪比（SNR）与特征参数联合估计的方法完成自动分类,仿真结果表明在SNR高于0dB时识别率大于95％。     

Blind Source Recovery in a State-Space Famework: Algorithms for Static and Dynamic Environments

This paper describes a generalized state space Blind Source Recovery (BSR) framework obtained by using the Kullback-Lieblar divergence as a performance functional and the application of optimization theory under the constraints of a feedforward state space structure. Update laws for both the non-linear and the linear dynamical systems have been derived for the domain of dynamic blind source recovery along both ordinary stochastic gradient and the Riemannian contra-variant gradient directions. The choice of the rich state space demixing network structure allows for the development of potent learning rules, capable of handling most filtering paradigms and convenient extension to non-linear models. Some particular filtering cases are subsequently derived from this general structure and are compared with material in the recent literature. Some of this reported work has also been implemented in dedicated hardware/software. An illustrative simulation example has been presented to demonstrate the online adaptation capabilities of the proposed algorithms.     

基于回归特征分析的BPSK信号盲处理结果可信性评估

针对二相编码（Binary phase shift k eying,BPSK）信号盲处理过程中信号调制方式识别及参数估计的可信性评估问题,提出了一种 基于线性回归特征分析的处理方法。根据特定处理结果建立参考信号,并将参考信号与原始 观测信号进行相关累加取模值,在分析相关累加模值曲线在原序与升序两种不同情形下线性回归 差异的基础上,分别定义了排序回归系数及分段回归符号两大特征,用以判别BPSK信号盲处 理结果的可信性。计算机仿真结果表明,在低信噪比、缺乏信号先验信息的条件下,本算法 可对BPSK信号盲处理结果的可信性进行有效检验。     

Linear State Space Feedforward and Feedback Structures for Blind Source Recovery in Dynamic Environments

Blind Source Recovery (BSR) denotes recovery of original sources or signals from environments, which may include convolution, temporal variation, and even non-linearity—without necessarily performing explicit environment identification. This paper presents two separate multiple-input multiple-output (MIMO) structures for dynamic linear Blind Source Recovery (BSR) of multiple stochastically independent source signals. We propose linear state space models for both the mixing environment and the demixing (or the recovering) adaptive network. The demixing network may assume either the feedforward or the feedback state space configuration. Separate algorithms have been derived for the adaptive estimation of the feedforward and the feedback demixing/recovering network’s parameters. These algorithms are based on the multivariable optimization theory and utilize the Riemannian contra-variant gradient (or the natural gradient) search under the constraints of a state space representation. Auxiliary conditions for the convergence of these algorithms have also been derived and discussed. Illustrative simulation examples have been included to demonstrate the successful adaptation results for both proposed demixing network structures in the case of an infinite impulse response (IIR)-type mixing environment for example source statistical distributions.     

Design of Nonuniform Linear Antenna Array Geometry and Signal Processing Algorithm for DOA Estimation of Gaussian Sources

Abramovich, Yuri I., and Spencer, Nicholas K., Design of Nonuniform Linear Antenna Array Geometry and Signal Processing Algorithm for DOA Estimation of Gaussian Sources, Digital Signal Processing10 (2000), 340–354.This paper discusses the problem of direction-of-arrival (DOA) estimation for Gaussian sources that have arbitrary correlation: from independent to fully correlated. For independent sources, the antenna array design is governed by two competing considerations: maximum aperture, which inclines toward increasing sparsity for a given number of array sensors, and identifiability, which tends to exclude extreme sparsity. For fully correlated sources, these two competing criteria are augmented by a third which allows for the initialization of DOA estimation by the generalized spatial smoothing (GSS) technique. The maximum number of fully correlated sources is in turn an important factor in the GSS algorithm and subsequent array geometry design. We present a geometry optimization technique that permits accurate DOA estimation of arbitrarily correlated sources.     

一种基于压缩感知和动态时间规整的信号肽特征提取新算法

准确识别出信号肽对蛋白质的研究和定位有着非常重要的意义。压缩感知技术能够在保留生物序列主要信息的同时降低冗余信息,将高维信息投影到低维空间上进行特征提取。因此本文基于压缩感知技术再结合动态时间规整算法提取出新的特征向量,提出一种高鉴别性的信号肽特征提取新方法。该算法所提取的特征不但体现了信号肽中的氨基酸组成、排列顺序、结构等重要信息,还能把信号肽的不同区域在时间维度中非线性地弯曲对整,为机器学习算法提供有效的信号肽特征表达。实验结果显示,新方法提取的特征向量在3个数据集Eukaryotes, Gram+ bacteria, Gram-bacteria上的识别率分别达到99.65%, 98.05%和98.56%,并且这种方法能简单地运用到其他生物序列的识别过程中。     

Cognitive Architecture for Robust Adaptive Control of Robots in a Team

The objective of this paper is to present a cognitive architecture thatutilizes three different methodologies for adaptive, robust control ofrobots behaving intelligently in a team. The robots interact within a worldof objects, and obstacles, performing tasks robustly, while improving theirperformance through learning. The adaptive control of the robots has beenachieved by a novel control system. The Tropism-based cognitive architecturefor the individual behavior of robots in a colony is demonstrated throughexperimental investigation of the robot colony. This architecture is basedon representation of the likes and dislikes of the robots. It is shown thatthe novel architecture is not only robust, but also provides the robots withintelligent adaptive behavior. This objective is achieved by utilization ofthree different techniques of neural networks, machine learning, and geneticalgorithms. Each of these methodologies are applied to the tropismarchitecture, resulting in improvements in the task performance of the robotteam, demonstrating the adaptability and robustness of the proposed controlsystem.     

Soft Computing Based Pattern Classifiers for the Obstacle Avoidance Behavior of Intelligent Autonomous Vehicles (IAV)

To ensure more autonomy and intelligence with real-time processing capabilities for the obstacle avoidance behavior of Intelligent Autonomous Vehicles (IAV), the use of soft computing is necessary to bring this behavior near to that of humans in the recognition, learning, adaptation, generalization, reasoning and decision-making, and action. In this paper, pattern classifiers of spatial obstacle avoidance situations using Neural Networks (NN), Fuzzy Logic (FL), Genetic Algorithms (GA) and Adaptive Resonance Theory (ART) individually or in combination are suggested. These classifiers are based on supervised learning and adaptation paradigms as Gradient Back-Propagation (GBP), FL, GA and Simplified Fuzzy ArtMap (SFAM) resulting in NN/GBP and FL as Intelligent Systems (IS) and in NN/GA, NN/GA-GBP, NN-FL/GBP and NN-FL-ART/SFAM as Hybrid Intelligent Systems (HIS). Afterwards, a synthesis of the suggested pattern classifiers is presented where their results and performances are discussed as well as the Field Programmable Gate Array (FPGA) architectures, characterized by their high flexibility and compactness, for their implementation.