520.735 Sensory Information Processing                                     Assignment #1 (due 02-22-01)

Reading:

 Read the Introduction and Chapter I -Discrete Noiseless Systems- from the Shannon and Weaver book or

the Shannon paper.

Project:

The point of this project is to characterize the natural stimuli in a sensory modality of you choice: vision, audition, touch,  e.t.c.

1) First you need to define the task or tasks in which you are interested. This could be as general as natural scenes or natural situations or as specific as discriminating between Braille. Contact one of the instructors if you are unsure.Be careful to select a modality for which data corpora exist; otherwise you may have to collect the necessary data so that you can do the remainder of the assignment.

2) Investigate and characterize the statistical properties associated with the data. For example, if you choose vision, might characterize the joint distributions associated with spatial frequencies, or temporal frequencies, or color, or intensities, orpolarization, or some combination of these.

3) Characterize the first-order probability distributions of your source -- that is, generate relative frequency distributions (histograms) for your data corpora. If we assume that the measurements are i.i.d., calculate the entropy of the source and the the entropy rate?

4) Characterize the second-order probability distributions of the source. How are the different values correlated? Depending on the nature of your task, you might want to develop a Markov model or characterize the power spectrum of your source. (Under what conditions are these equivalent?)

5) Report on the sensors that were employed to collect these data?