西安工业信号检测与估计SDE_01 CourseIntro.pptVIP

05-02-29 SDE_01 introduction 信号检测与估值Theory of Signal Detection and Estimation 雷斌QQ leibin@ gropemind@163.com What is SDE About? Goal目标: Suck useful information out of messy data 从杂乱的数据中提取有用 Strategy策略: Formulate probabilistic model of data y, which depends on underlying parameter(s) ? 阐明y,与未知参数? 的相关程度 Terminology depends on parameter space: 参数空间的相关术语 Detection (simple hypothesis testing):检测（简单假设检验） ? ?{0,1}, i.e. 0=target absent, 1=target present 目标存不存在 Classification (multihypothesis testing): 分类（多假设检验） ? ?{0,1,…,M}, i.e. ? ?{DC-9, 747, F-15, MiG-31}目标类型 Estimation 估值 ? ?Rn, Cn, etc. (not too hard) ? ?L2(R), (square-integrable functions), etc. (harder) Signal-to-Noise Ratio信噪比 Measured in dB!!!!! Ability to recognize target in random noise Noise is always present. At some range, noise is greater that target’s return Noise sets lower limit of unit’s sensitivity Threshold level 阈值 used to remove excess noise 课程描述Course Description 课程简介 包含 估计理论 Estimation theory 从噪声中接收到的信号里提取信息的统计参数估计 利用样本资料所提供的信息，估计总体分布中的某些未知参数叫：参数估计。 分点估计与区间估计两类。 和 检测理论 Detection theory 噪声中信号检测的统计假设检验理论 简单假设检验－一元假设检验 Yes or Not 多元假设检验 ：A/B/C/D…which one 估计理论 Estimation theory 估计理论 Estimation theory (con) 检测理论 Detection theory or  信号检测理论signal detection theory 检测理论 Detection theory (con) To be B or not to be B:That is the Bayesian question To be B 贝叶斯估计：通过随机变量的先验概率推测待估参数A Bayesian analysis treats推测 ? as a random variable with a “prior” density 先验概率p(?) Data generating machinery is specified by a conditional density p(y|?) 信号发生装置依照条件概率产生数据 Gives the “likelihood似然” that the data y resulted from the parameters ? Inference usually revolves around the posterior density, derived from Bayes’ theorem: Tasks of Statistical Signal Processing统计信号处理的任务 建立数据的统计模型 Create statistical model for measured data 用现有的理论推导 Find fundamental limitations on our ability to perform inference on the data Cramèr-Rao bounds, Chernov bounds, etc. 找出最佳或准最佳的估计 Develop an optima