Academics
  /  
Courses
  /  
Descriptions
ELEC_ENG 395: Adaptive Signal Processing and Learning

Quarter Offered

None : TuTh 11:00am-12:20pm ; Honig

Prerequisites

ELEC_ENG 202, ELEC_ENG 302

Description

CATALOG DESCRIPTION: discrete-time random process, second-order statistics, autoregressive and moving average processes, linear prediction, Wiener filter, stochastic gradient (Least Mean Square) algorithm, least squares estimation, introduction to Kalman filter. 

REQUIRED TEXT: S. Haykin, "Adaptive Filter Theory", Prentice-Hall, 2013. 

COURSE DIRECTOR: Prof. Mike Honig

COURSE GOALS: To provide an introduction to adaptive signal processing methods with applications to compression, prediction, model estimation (learning), and array processing. 

PREREQUISITES BY COURSES: 202, 302 

PREREQUISITES BY TOPIC: 

        ITEM 1: Probability     

        ITEM 2: Frequency-domain (spectral) analysis 

        ITEM 3: Familiarity with z-transforms. 

COURSE TOPICS: 

  1. Applications: speech compression, financial forecasting, array processing
  2. Discrete-time random process, second-order statistics, filtering
  3. Autoregressive and Moving Average processes
  4. Linear prediction, Wiener filter
  5. Gradient and stochastic gradient (Least Mean Square) algorithms
  6. Least squares estimation and filtering
  7. Introduction to Kalman filter 

GRADES: A weighted combination of homework, midterm, and project. 

COURSE OBJECTIVES:  When a student completes this course, s/he should be able to: 

  1. Characterize a wide-sense stationary discrete-time random process in terms of second-order statistics and spectral desnity.
  2. Model a given signal or time-series as an AR, MA, or ARMA random process.
  3. Compute the optimal (Wiener) predictor or filter from second-order input statistics.
  4. Design a stochastic gradient algorithm to satisfy particular performance criteria.
  5. Compute a Least Squares approximation of the Wiener filter from measurements.
  6. Simulate adaptive signal processing algorithms to compare relative performance.
  7. Apply a state-space model and the Kalman filter to solve a basic tracking problem.