A method minimizing the correlation properties of a noise field and improving detection of signals embedded in anisotropic noise

This paper deals with the development of a processing technique that improves the signal-to-noise ratio (SNR) at the single sensor for a received signal that is embedded in a partially correlated noise field. The approach of this study is unique in that the noise is treated as being non-white and partially correlated. The concept of the proposed development is based on the time interval over which the temporal coherence or correlation properties of a noise field are defined. For narrowband signals, the associated temporal coherence period is much longer than the correlation time interval of the anisotropic noise field. Thus, a coherent integration of discontinuous segments of received signals will enhance the SNR at the single sensor by lowering the correlation properties of the associated non-white noise. Reconstruction of the narrow-band signal time series, with improved SNR at the sensor will allow the use of the existing high resolution techniques to be utilized more effectively by lowering their threshold values in order to detect very weak signals. The intention here is to integrate the characteristics of the real anisotropic noise field during the preliminary processing stages of the received signals by an array of sensors. Simulations show that the proposed method can be integrated in the signal processing functionality of sonar and radar systems