Communications-centric Motivation: Communication over noisy channels.
Noise
Source Encoder Channel Decoder Destination
Examples: Email over internet, Voice over phone network, Data over USB storage. In these examples, the source would be the voice, email text, and data, the channel would be the (wired or wireless) phone network, internet, and storage device. Purpose: Reproduce source data at the destination (with presence of noise in channel) Channel can introduce noise (static), interference (multiple copies at destination arriving at different times from radio waves bouncing on buildings, etc), or distortion (amplitude, phase, modulation) Sources have redundancy, e.g. the english language, can reconstruct words from parts: th_t that, and images, where individual pixel differences are not noticed by the human eye. Some sources can tolerate distortion (images). A typical example of an encoder is broken into three components: From Source Source Encoder Channel Encoder Modulator to Channel The source encoder outputs a source codeword, an efficient representation of the source, removing redundancies. The output is taken to be a sequence of bits (binary) The channel encoder outputs a channel codeword, introducing controlled redundancy to tolerate errors that may be introduced by the channel. The controlled redundancy is taken into account while decoding. Note the distinction from redundancies present in the source, where the redundancies may be arbitrary (many pixels in an image). The codeword is designed by examining channel properties. The simplest channel encoder involves just repeating the source codeword. The modulator converts the bits of the channel codeword into waveforms for transmission over the channel. They may be modulated in amplitude or frequency, e.g. radio broadcasts. The corresponding decoder can also be broken into three components that invert the operations of the source encoder above:
From Channel Demodulator Channel Decoder Source Decoder To Destination
We call the output of the demodulator the received word and the output of the channel decoder the estimated source codeword. Remark. Later we will see that separating the source and channel encoder/decoder as above does not cause a loss in performance under certain conditions. This allows us to concentrate on each component independently of the other components. Called the Joint Source Channel Coding Theorem.