Turbo Codes and Channel Estimation in Wireless Systems
This thesis focuses on practical implementation issues of turbo codes with imperfectly estimated fading wireless channels and limited code length.
Channel estimation can be divided into estimation of static parameters, such as constant timing and phase offset, and dynamic parameters, such as fading channel amplitudes. In this thesis, it is shown that the static parameter estimation can be performed using only the knowledge of the code structure without transmitting any additional pilot symbols. The non-linear phenomena, such as cycle slips and phase ambiguity can also be resolved using only code properties.
The dynamic channel estimation is more complex and requires the use of known pilot symbols (Pilot Symbol Assisted Modulation - PSAM) to initialize the decoding. This thesis quantifies the influence of different pilot symbol structures and provides tight bounds on channel estimation errors with Wiener filtering. A similar method, using a pilot symbol matrix and a Wiener filter, is shown to improve the performance of the Orthogonal Frequency Division Multiplexing (OFDM) system as well. A new method which modifies the convergence properties of joint channel decoding and estimation is shown to provide asymptotically lower bit error rate.
Finally, a new class of codes using a frequently terminated first trellis is presented. This class of codes can easily be implemented with fixed frame lengths and provides almost linearly changing spectrum of rates allowing easy adaptation to the changing channel signal to noise ratio. Moreover, the properties of these codes can be used to improve channel estimation and allow for simple implementation of unequal error protection (UEP).
unequal error protection