Zeit-Frequenz-Verarbeitung von Zufallsprozessen

Summary: The research performed within this project addressed several areas of time-frequency signal processing and wireless communications. Statistical time-frequency methods were developed in the fields of nonstationary signal estimation and detection; knock detection for car engines; speech enhancement; nonstationary spectral analysis; time-varying filtering; simulation, sounding, estimation, and prediction of mobile radio channels; and multiple-antenna communications. The project resulted in 8 book contributions, 5 journal articles in reviewed international journals, and 15 papers at international conferences; it also led to the formulation of a new research project. In the course of the project, informal contacts with the Austrian company Siemens Austria AG were maintained. The results obtained within the project can be grouped into three areas as discussed in the following. Time-frequency methods for nonstationary statistical signal processing We developed a time-frequency formulation, time-frequency design, and time-frequency implementation of nonstationary signal estimators and signal detectors. We applied our time-frequency estimation and detection methods to speech enhancement and knock detection for car engines. We also developed various extensions and a unified formulation of existing time-varying spectra, as well as a novel class of time-frequency coherence functions. These results were applied to the spectral and coherence analysis of nonstationary pressure and vibration signals measured in car engines. Time-frequency methods for time-varying filters and mobile radio channels Time-varying systems and filters are important in many signal processing applications. We developed a time- frequency transfer function calculus for time-varying systems. This calculus was applied to the design and efficient online implementation of time-varying filters and to an analysis of the measurement errors of practical mobile radio channel sounders. We also developed time-frequency methods for estimating the scattering function of mobile radio channels and a computationally efficient channel simulator for mobile radio channels with arbitrary scattering function (delay-Doppler profile). Signal processing for wireless communications Wireless communications is a field of tremendous practical importance that allows a natural application of time- frequency methods. This is particularly true for a wireless modulation scheme known as orthogonal frequency division multiplexing (OFDM). We developed an OFDM receiver that is based on predictive estimation and equalization of time-varying wireless channels. This receiver needs no training symbols and thus allows for increased data rates. Furthermore, we investigated intersymbol/intercarrier interference effects in OFDM systems that are due to time-varying channels. Finally, we developed a technique for blind equalization of time-varying channels and we extended this technique to OFDM communication systems using multiple transmit and receive antennas.