From seed
Jump to navigation Jump to search

Adaptive Signal Processing[edit]

Computer systems are increasingly receiving their input from sensors such as microphones, video cameras, range finders etc. Digital signal processing provides the foundational methodology for understanding and processing such signals. Adaptive signal processing methods change their parameters so as to optimize their performance with respect to the observed signals.

This course is an introduction to digital signal processing and adaptive signal processing intended graduate students with limited prior exposure to the subject. Prior background required is linear algebra and (helpful but not necessary) complex numbers.

Homework for the course will consist of exercises implementing and testing algorithms in matlab. There will be a class project analyzing real-world data and a written final exam.


Days: TuTh Time: 11:00a - 12:20p Room: EBU3B 4217

Class files[edit]


  • Linear filters, Impulse Response, The frequency domain, Fourier transform.
  • A review of complex numbers.
  • Z transform, the Phase Diagram, poles and zeros, analysis of stability, Phase diagrams of linear systems.
  • Noise cancellation.
  • Adaptive Linear filters, Principle of orthogonality, Least Mean Square (LMS), Recursive Least Squares (RLS), levinson Durbin Recursion.
  • Echo cancellation.
  • Linear Prediction codes and vector quantization in vocoders.
  • Markov Models, Hidden Markov Models, inference of Latent variables.
  • MFCC, speech recognition using HMM and mixture of gaussians.
  • Kalman Filters, Extended Kalman Filters. Particle Filters, Tracking using online learning.
  • Beam-forming.

Recommended books[edit]