Signal Processing Technique Based On The Separability Characteristics In Autocorrelation Domain

One of the fundamental tasks of signal processing is to detect and estimate theuseful signal from the noise and interference, since the observed signal is always mixedwith the interference and noise. And to finish this taks, the separability of signal is oneof the key problems. Usually, the signal is analysised and processed in time domain orfrequency domain, and it is difficult to separate the signal from the interference andnoise when the signal is overlapping both in the time domain and in the frequencydomain. Hence, it is necessary to find the other domain which can be used to separatethe signals. The autocorrelation domain is another signal processing domain, and thewhite noise is naturally can be separated from the other non-white signals due to itsunique characteristic in this domain. This paper focus on the separability of signal inautocorrelation domain, and the signal processing techniques based on the characteristicof the singal which can be used to separate the singal from the interference, noise in theaucorrelation domain. The research works of this thesis are as following:1. Study on the signal description in the autocorrelation domain, and uncover theseparability of signal in the autocorrelation domain. The definition of separability inautocorrelation domain of signals and signal subsets are given in the paper; it can beused to be a criterion for characteristic extracting and signal designing in theautocorrelation domain.2. When the signals satisfy the separability requirement in autocorrelation domain,and the signal subset also can be separable from the interference subset, theautocorrelation filter is discussed, and the channel estimation method based on theseparability characteristics in autocorrelation domain is presented. The simulationresults indicate that the methods based on the characteristics in autocorrelation domaincan effectively achieve interference cancellation and channel estimation.3. When the signal and the interference exit nonzero autocorrelation value atnonzero time-lags, an improved MMSE estimator is proposed. It is constructed based onthe conventional MMSE estimator and the nonzero time-lags characteristics of signal inautocorrelation domain. The simulation results indicate that the improved estimator can use the second-order statistics at nonzero time-lags from both signal and interference toreduce the estimate error.4. Based on the signal processing techniques which proposed in this thesis, asignal detection and estimation scheme for wireless communication system is proposed.The channel estimation method and autocorrelation filter based on signal separabilitycharacteristics in autocorrelation domain are used to achieve channel estimation andinterference cancellation, and then use the improved MMSE estimator to estimate theuser signal accurately.5. An improved winier filter based on the autocorrelation characteristic at nonzerolags is presented to get the better filter results. The validity of the methods researched inthis paper is also verified by experiment on the actual audio signals.