What will you get out of this course?

A comprehensive view of modern communications! First we turn everything into bits!

• Quantization
• Level Quantization 
• Projection onto basis then quantization
$$s(t) \approx \sum_{i=1}^N a_i \phi_i(t)$$




where $a_i = \int s(t) \phi_i(t) dt$

FEATURES: finite basis set, then quantize the a_i to some number of bits.

• Extraction of system parameters then quantization

Linear system model of source (like vocoder) and send only system parameters and exicitation.
• Get as creative as you'd like!
• BUT! at heart there is the fundamental issue of how many bits you need to represent a given time waveform which is just one of many (how many?) possible (put your hands together in prayer because this is a solemn concept) MESSAGES.
• Onward to bit processing!

 

 



• Remove redundancy (source coding, an Information Theory concept [332:542])
• We code for the channel (channel coding)
• We do something to this code and produce a waveform (modulation). Oh, I've got a bone to pick with the ``old-heads'' here... more later.
• Then we send (transmit) the lovely waveform to be corrupted by the nasty channel.
• We take our bruised waveform in at the receiver and invert the modulation Again, bones to pick (later)!
• Now we get fancy and do special decoding at the channel and source level to produce a good estimate of what the bits were.
• The goodness criterion is how often we screw up a bit estimate (either of the source, or the source code) $G(x(n),\hat{x}(n))$or $G(b(n),\hat{b}(n))$
• BIT ERROR RATE IS CLEAN and NON-FUZZY

 

 

(but again caveat emptor!!!!!)

1999-01-20