One way to correct errors, assuming that errors are infrequent, is to encode the same information several times, in short bursts. For instance, suppose we want to send the following bit string
|1001|1100|0000|0101|We could instead send two backup copies of each group of N bytes, say N=4:
| | | | | |1001|1001|1001|1100|1100|1100|0000|0000|0000|0101|0101|0101| | | | | |Suppose an error occurs in transmission, so that this message is received as:
| | | | |
|1001|1001|1001|1100|1100|1100|0010|0000|0000|0101|0101|0101|
| | | ^ | |
| (error)
This error can be corrected by assuming that the most probable
value of each bit is chosen by majority decision. i.e. it is
unlikely that the same error would occur in the two backup
copies.
This approach is used essentially by CD players with read-ahead.