2 After presenting the image models, characters and existing methods of blind image deconvolution, Iterative Blind Deconvolution(IBD) algorithm Maximum Likelihood (ML) algorithm, Richard-Lucy (RL) algorithm, Nonnegativity and Support Constrants Recursive Inverse Filtering (NAS-RIF) algorithm are analyzed and realized, then based on a lot of experiments the algorithms are compared with each other.

2、介绍了图像盲解卷积的数学模型、性质及现有的方法之后,深入分析并实现了迭代盲解卷积(Iterative Blind Deconvolution, IBD)方法、最大似然(Maximum Likelihood, ML)方法、Richard-Lucy(RL)方法、具有非负性和支持域约束条件的递归逆滤波(Nonnegativity and Support Constrants Recursive Inverse Filtering, NAS-RIF)算法,做了大量试验,比较了各类算法的性能。

Theoretical analysis and experimental result indicates that the transformation can reduce the miss recognition rate up to 8.6%, compared to Maximum Likelihood (ML) transformation which is commonly used in GMM.

A rapid timing acquisition algorithm,which is based on the maximum likelihood(ML) criterion,is developed for UWB(ultra-wideband) communications based on designed training symbols.

Vrarance and covarance components were estimated by a restricted maximum likelihood (REML)approach for the 6 traits, and heritability estimates of which were 0.24, 0.215, 0.29,0.48 and 0.36 resp.

用约束最大似然估计法(REML:Restricted maximum likelihood)估计方差组分,得到以上6性状的遗传力分别为0.24,0.215,0.29,0.48和0.36。

Maximum Likelihood (ML) frame synchronization estimator and its corresponding simplified estimators are presented.

论文设计了一种采用最大似然估计(ML)帧同步估计器和相应的简化估计器。

The Maximum likelihood(ML) decoding of the proposed codes can be implemented by the sphere decoder at a moderate complexity and the complexity of the decoding algorithm is independent of signal constellation size.

The theory analysis and result of experiment indicates that the transformation can reduce the error recognition ratio to 2.0% ,comparing to Maximum Likelihood(ML) transformation which is mostly used in GMM.

By comparing the CD-RELAX with the Alternating Projection of the Maximum Likelihood(ML-AP) , the opinion of the exact fitting and the non-exact fitting was proposed .

The relative efficiencies of moment estimators as compared with the maximum likelihood and the stepwise estimators are computed.

Generalized Maximum Likelihood Algorithm for Direction-of-Arrival Estimation of Coherent Sources

The generalized maximum likelihood (GML) algorithm for direction-of-arrival estimation is proposed.

Secondly, the comparison between the GML algorithm and the conventional deterministic maximum likelihood (DML) algorithm is presented based on their respective geometrical interpretation.

In this article, we developed a new algorithm for eigenspace-based adaptation restricting eigenvoices in clustered subspaces, and maximum likelihood (ML) criterion was replaced with maximum aposteriori (MAP) criterion for better parameter estimation.

The non-linear estimation of time-varying parameters of singals in the presence of noise is discussed. A concept of instantaneous maximum likelihood estimation is presented and is shown to be asymptotically sufficient. A new approach of optimum estimation has been found by which the diffcult problem of non-linear estimation of time-varying parameters of signals is simplified as first finding the instantaneous maximum likelihood estimation and then its optimum processing.The apparatus used to realize...

The non-linear estimation of time-varying parameters of singals in the presence of noise is discussed. A concept of instantaneous maximum likelihood estimation is presented and is shown to be asymptotically sufficient. A new approach of optimum estimation has been found by which the diffcult problem of non-linear estimation of time-varying parameters of signals is simplified as first finding the instantaneous maximum likelihood estimation and then its optimum processing.The apparatus used to realize the instantaneous maximum likelihood estimation is defined as optimum demodulator and its general configuration is given. The application of the above theory is described with the phase-modulated signal used as an example. Finally, the results obtained in this paper are compared with those of D. C. Youla and D. L. Snyder.

It is agreed by most statisticians that in determing the LD_x of a drug the probir method would give a satisfactory solution. However, the computation required for probit analysis with maximum likelihood method is tedious and time consuming. Research workers are clearly justified in pressing statisticians to invent simpler methods of analysis.This article presents an alternative method of probit analysis for determing LD_x, the "weighted approximation method". This is much simpler than the method of maximum...

It is agreed by most statisticians that in determing the LD_x of a drug the probir method would give a satisfactory solution. However, the computation required for probit analysis with maximum likelihood method is tedious and time consuming. Research workers are clearly justified in pressing statisticians to invent simpler methods of analysis.This article presents an alternative method of probit analysis for determing LD_x, the "weighted approximation method". This is much simpler than the method of maximum likelihood and is a good approximation to the latter.Let x_1, x_2, …, x_k be the k logarithm doses with equally spaced I; p_1, P_2,…, P_k, the corresponding percentage of positive responses; y_1, y_2,…, y_k, the probits and w_1, w_2,…, w_k, the weights of corresponding probits, in which y and w may be obtained from the tables listed in the text.With the analysis of weighted linear regression, the slope of weighted regression lineb= CA/IB (1) then and the standard deviation of M (2)there C=1 for an odd number of doses, C=2 for an even number of doses, the calculation of A and B are tabulated as follows:With polynomial coefficients λ of linear regression, formulae (1) and (2) may be for an arbitrary number of doses.An example is described which describes that the weighted approximation method has a negligible error, while the simplified probit method already recommended by Gu Hanyi has an unnegligible error as compared with the maximum likelihood method.A criticism of "simplified probit method" is presented in the appendix.

Based upon the method of statistical testing,this paper analyses and compares the methods of moments, maximum likelihood and curve-fitting in the case of Pearson type Ⅲ distribution. For the curve-fitting method, a comprehensive comparative analysis is made on seven criteria of curve fitting and mainly on the two of them-the least squares criterion φ1 andthe least absolute-values criterion φ2, as well as on the various plotting position formulas.Moreover, the role of historical flood is also examined....

Based upon the method of statistical testing,this paper analyses and compares the methods of moments, maximum likelihood and curve-fitting in the case of Pearson type Ⅲ distribution. For the curve-fitting method, a comprehensive comparative analysis is made on seven criteria of curve fitting and mainly on the two of them-the least squares criterion φ1 andthe least absolute-values criterion φ2, as well as on the various plotting position formulas.Moreover, the role of historical flood is also examined. The results of calculation indicate that the quantity Xp derived from the moment method has a much smaller expected value EXP than the true value xp?of the population and that this is on the unsafe side. The maximum likelihood method yields somewhat better result but it cannot be accepted as a universal method because it yields no solution to Pearson Ⅲ .curve when Cs≥2. For the curve fitting method, it is found that the result is substantially affected by the criterion used for curve fitting.In general, φ2 criterion associated with Weibull formula ( abbr. in φ2-W ) may give slightly greater EXP than xpo and smaller root-mean-square error σXp. Consequently, φ2-W curve-fitting method proves to be the better one among all the methods available. The use of historical flood would usually increase EXP and decrease σXp and is, therefore, an important means to improve the computational results.