Computer Analysis and Modeling of Biological Signals and Systems

Spring term 2022

Instructors: Mark Liberman
Stephen Isard

(Hybrid) Lecture/Discussion: MW 12:00-1:30

Discussion via Piazza

Software: Examples and homework exercises will require Octave or MATLAB.

Consult the schedule, which also has links to lecture notes.

Homework: There will be about 6 homework assignments, plus a term project.
Submit assignments via the course Canvas site.
Structure: There will be a series of homework exercises using the computer language Octave (the free version of MATLAB), and a term project, which can be a summary of existing methods and results in some area of interest to you, or can be new work of your own. Grading will be based on the homework (40%), the term project (50%), and class participation (10%).
Prerequisites: Digital signal processing is mainly applied linear algebra. There are also basic connections to calculus and probability, and the physics of signals will also come up. The course will review the needed mathematical concepts, but if they are all entirely new to you, you will have to work hard to learn both the basic mathematics and its application. However, a genuine interest in understanding, modeling, or mimicking biological systems will go a long way. In the past, participants who have found the course worthwhile had backgrounds ranging from an MS in math to "nothing past high school algebra."
Texts and readings:

Extensive on-line lecture notes will be provided (linked to the syllabus).

Course Philosophy: Digital Signal Processing is basically a simple topic, whose fundamentals are easier for most people to understand than first-year college calculus is. It provides essential conceptual and practical tools for research in areas such as computer vision, phonetics and speech processing, neuroscience, computer music, and any other discipline that is concerned with the production, perception or interpretation of physical signals by living creatures.

However, DSP is usually taught to electrical engineers after three or four semesters of prerequisites; and then an EE DSP course usually includes some things that are not crucial for a biologically-oriented audience, while leaving out some other other things that are. A properly designed one-semester lab course can give interested students the foundation needed to understand and use DSP concepts and techniques in biological applications. This can be done without requiring a mathematical background much beyond basic algebra.

This course was developed in the mid 1990's by Mark Liberman and Eero Simoncelli. Eero now teaches a modified version, called Mathematical Tools for Neural Science, as a required course in the Neuroscience program at NYU.