This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: ECE 534: Elements of Information Theory, Fall 2010 Homework 10 SOLUTIONS all by Davide Basilio Bartolini Ex. 10.1 Text Onebit quantization of a single Gaussian random variable . Let X N (0 , 2 ) and let the distortion measure be squared error. Here we do not allow block descriptions. Show that the optimum reproduction points for 1bit quantization are radicalBig 2 and that the expected distortion for 1bit quantization is  2 2 . Compare this with the distortion rate bound D = 2 2 2 R for R = 1. Solution If we are using a 1bit quantization, we are allowed to use up to two reproduction points (say, p and p + ) for which two reconstruction regions are to be defined. It is quite straightforward that the optimal reconstruction regions in this case are the positive and negative axes; the reconstruction points may be found as the expectation for each over the given Gaussian distribution. For the positive axis: p + = integraldisplay x 2 2 2 e x 2 2 dx = integraldisplay 2 2 2 2 e y dy parenleftbigg substituting y = x 2 2 2 dy = 2 x 2 2 dx = x 2 dx parenrightbigg = integraldisplay radicalbigg 2 e y dy = bracketleftBigg radicalbigg 2 e y bracketrightBigg = radicalbigg 2 1 and, by symmetry, p = radicalBig 2 . Now we can compute the expected distortion D as the expected squared error: D = E f ( x ) [( X X ) 2 ] = integraldisplay parenleftBigg x + radicalbigg 2 2 parenrightBigg 2 1 2 2 e x 2 2 2 dx + integraldisplay parenleftBigg x radicalbigg 2 2 parenrightBigg 2 1 2 2 e x 2 2 2 dx = 2 integraldisplay ...
View
Full
Document
This note was uploaded on 01/19/2012 for the course ECE 534 taught by Professor Natashadevroye during the Fall '10 term at Ill. Chicago.
 Fall '10
 NatashaDevroye

Click to edit the document details