Abstract :
Summary form only. Wireless broadband communications systems are characterized by very dispersive channels. Channel estimation is therefore a crucial task for reliable wireless transmissions. In most practical systems, this task is carried out using training. In the case of slow-varying fading channels, a training sequence is usually appended at the beginning of the data burst (i.e. a preamblebased training). In the case of fast-varying channels, the pilot symbols are inserted in the data stream (a technique called pilot symbol assisted transmission [PSAM]). The pilot symbols are orthogonally multiplexed with data, either in time (such as in GSM), frequency (such as in OFDM) or code (such as in W-CDMA). An alternative and promising approach to preamble-based training and PSAM is superimposed training (ST). This scheme saves valuable bandwidth at the expense of a reduction in the information signal-to-noise ratio. In this talk, I will give a survey of all these training-based techniques. But in particular I will concentrate on the very recently published developments of the new paradigm Data Dependent Superimposed Training (DDST), and its latest applications in time-varying environments, MIMO channels and turbo equalisation.
