Scribe Notes 7 (by Xixuan Wu)
Generally source coding reduce the redundancy, channel coding add the redundancy.
Description of an arbitrary real number requires an infinite number of bits. Finite representation of a continuous random variable can never be perfect. So it's necessary to define the "goodness" of a representation of a source. The problem is: Given a source distribution and a distortion measure, what is the minimum expected distortion achievable at a particular rate? or What's the minimum rate required to get a proper distortion?
1. Quantization
The representation of X is denoted as .
1.1 1bit random variable Quantization Example
Let , assume the distortion measure is . We have only one bit to represent X, so it's clear that the bit will show whether X>0 or not. To minimize squared error, each reproduced symbol should be the conditional mean of its region, the equation and figure is showed as below:
The reconstruction points are at -0.79 and 0.79.
1.2 2bit Quantization Example
Like
Comments (0)
You don't have permission to comment on this page.