The transmission of information over a communication channel often involves channels whose characteristics are unknown. The information is therefore distorted by an unknown channel which in addition adds random noise to the message. If the channel characteristics are known, then the non random part of the distortion can usually be compensated at the receiver, and the problem is one of detecting the messages buried in noise. If the channel characteristics are unknown, then a crucial step is to estimate the channel first. Under certain conditions it is indeed possible to do this estimation from the received data itself. This approach, called blind identification of channels, has been studied in the past. The investigator studies new methods for such identification based on frequency domain information of the channel.
In recent years the tremendous power of redundant filter bank precoders has been established. Precoders not only allow linear distortions introduced by known channels to be cancelled with finite duration filters at the receiver, they also allow the identification of unknown channels without training sequences and without the need for fourth order statistics. The research studies several new methods for blind identification and channel equalization based on filter bank precoders. One of these is blind channel identification directly in the frequency domain. This offers tremendous simplicity in computations. The research also seeks to explore tradeoff of complexity between transmitters and receivers. In the past, precoders have emphasized low transmitter complexity of zero-padded systems. The investigators study zero-jamming at the receiver as an alternative. The research also addresses a number of important issues which arise in the context of designing systems that are resistant to channel nulls.
