基于HMM模型的煤炭开采冲击模型分析

针对煤炭生产开采中冲击地压的安全问题,提出基于隐马尔可夫(HMM)模型的煤炭开采冲击模型评估分析,通过选取煤炭开采环境正常以及异常状态下的采集数据样本作为离线训练样本,得到预开采环境模型参数;再通过实时检测煤炭开采采集数据样本作为在线训练样本,将实时数据送入预开采环境模型进行评估,在最大似然度下学习得出其最大概率,通过匹配得到系统当前运行状态。实验结果表明,该模型在煤炭开采冲击中有较高的可靠性,煤炭安全监测工作稳定。

几种准正交空时分组码的编码方案分析

讨论了几种准正交空时码的编码及解码问题;分析了准正交空时码编码矩阵的缺秩问题及最大似然解码时至少有两个符号相关联的现象,即分析了准正交空时码的分集增益减小特性与最大似然解码复杂度的增加问题;阐释了通过星座图旋转固然可以保证准正交空时码的分集增益但却无法克服其最大似然解码复杂度的增加,因此,必须从编码矩阵本身出发来解决算法复杂度,为研究以"面向单符号"的最大似然解码算法奠定基础。最后给出了一定的计算机模拟结果以说明理论分析的有效性。

Gaussian mixture model clustering with completed likelihood minimum message length criterion

An improved Gaussian mixture model （GMM）- based clustering method is proposed for the difficult case where the true distribution of data is against the assumed GMM. First, an improved model selection criterion, the completed likelihood minimum message length criterion, is derived. It can measure both the goodness-of-fit of the candidate GMM to the data and the goodness-of-partition of the data. Secondly, by utilizing the proposed criterion as the clustering objective function, an improved expectation- maximization （EM） algorithm is developed, which can avoid poor local optimal solutions compared to the standard EM algorithm for estimating the model parameters. The experimental results demonstrate that the proposed method can rectify the over-fitting tendency of representative GMM-based clustering approaches and can robustly provide more accurate clustering results.

Zadeh的隶属函数对似然方法、语义通信和统计学习的意义

流行的似然方法不合适数据先验分布(即信源)可变场合。为此,我们把Zadeh的隶属函数看做预测模型,用隶属函数和可变信源产生似然函数,用平均对数标准(normalized)似然度定义语义信息测度。这样可以保证:(1)坚持使用最大似然准则;(2)预测模型适合信源可变场合;(3)得到的语义贝叶斯预测兼容贝叶斯定理;(4)预测模型能表达语义,便于理解。一组隶属函数构成一个语义信道,优化隶属函数就是使语义信道匹配Shannon信道,产生多标签模糊分类。文中介绍了通过两种信道相互匹配求解最大似然度的迭代算法。几个例子显示这种算法用于检验、估计和混合模型时,收敛快速且可靠。

Blind source separation based on generalized gaussian model

Since in most blind source separation(BSS)algorithms the estimations of probability density function(pdf)of sources are fixed or can only switch between one sup-Gaussian and other sub-Gaussian model,they may not be efficient to separate sources with different distributions.So to solve the problem of pdf mismatch and the separation of hybrid mixture in BSS,the generalized Gaussian model(GGM)is introduced to model the pdf of the sources since it can provide a general structure of univariate distributions.Its great advantage is that only one parameter needs to be determined in modeling the pdf of different sources,so it is less complex than Gaussian mixture model.By using maximum likelihood(ML)approach,the convergence of the proposed algorithm is improved.The computer simulations show that it is more efficient and valid than conventional methods with fixed pdf estimation.

A novel estimation algorithm for torpedo tracking in undersea environment

A novel estimation algorithm is introduced to handle the popular undersea problem called torpedo tracking with angle-only measurements with a better approach compared to the existing filters. The new algorithm produces a better estimate from the outputs produced by the traditional nonlinear approaches with the assistance of simple noise minimizers like maximum likelihood filter or any other algorithm which belongs to their family. The introduced method is extended to the higher version in two ways. The first approach extracts a better estimate and covariance by enhancing the count of the intermediate filters, while the second approach accepts more inputs so as to attain improved performance without enhancement of the intermediate filter count. The ideal choice of the placement of towed array sensors to improve the performance of the proposed method further is suggested as the one where the line of sight and the towed array are perpendicular. The results could get even better by moving the ownship in the direction of reducing range. All the results are verified in the MATLAB environment.

Study on Probability Distributions of Multi-Timescale Aerosol Optical Depth Using AERONET Data

The probability distribution analysis is per-formed for multi-timescale aerosol optical depth (AOD) using AErosol RObotic NETwork (AERONET) level 2.0 data.The maximum likelihood estimation is employed to determine the best-fit probability density function (PDF),and the statement that the fitting Weibull distribution will be light-tailed is proved true for these AOD samples.The best-fit PDF results for multi-site data show that the PDF of AOD samples with longer timescale in most sites tends to be stably represented by lognormal distribution,while Weibull distribution is a better fit for AOD samples with short timescales.The reason for this difference is ana-lyzed through tail characteristics of the two distributions,and an indicator for the selection between Weibull and lognormal distributions is suggested and validated.The result of this research is helpful for determining the most accurate AOD statistics for a given site and a given time-scale and for validating the retrieved AOD through its PDF.

Effect of the span of Australian GPS coordinate time series in establishing an optimal noise model

The span of coordinate time series affects the determination of an optimal noise model. We analyzed position data recorded for 10 continuous Global Positioning System （GPS） sites from 1998.0 to mid-2009 on the Australian Plate to estimate the best noise model and thereafter obtain the true uncertainties of the velocity, employing the maximum likelihood estimation （MLE） method. MLE was employed to analyze the data in four ways. In the first two analyses, the noise was assumed to be a combination of flicker noise and white noise for the raw time series and spatially filtered time series. In the final two analyses, the spectral indices and amplitudes were simultaneously estimated for a power law noise plus white noise model for the raw time series and spatially filtered time series. We conclude that the noise model of GPS time series in Australia can be best described as the combination of flicker noise and white noise. Velocity uncertainties fall below -0.2 mm/yr when the time span exceeds -9.5 years. A comparison of noise amplitudes and maximum likelihood estimation values between the raw and spatially filtered time series suggests that traditional spatial filtering to remove common-mode errors might not be applicable to the raw time series of this region.