On the feasibility of totally blind channel identification in communication systems using guard intervals

Blind identification techniques estimate the impulse response of a channel by exploiting known finite alphabet or statistical properties of the transmitted symbols. Alternatively, oversampling the output is known to introduce dependencies also exploitable for channel identification. This paper proves the feasibility of estimating the channel by relying instead on the short sequences of zeros, known as guard intervals or zero padding, introduced between transmitted blocks by a number of communication protocols. Since no property of the transmitted information symbols is assumed, the method is called totally blind channel identification. It is proved that totally blind channel identification requires only two received blocks to estimate the channel.