A Novel Normalization Algorithm Based on the Three-Dimensional Minimum Variance Spectral Estimator

In active sonar and radar the noise power in the presence of reverberation and clutter is not stationary. This makes automated detection, tracking, and classification of targets difficult. One way to deal with this problem is to normalize the data. The goal of normalization is to equalize the noise power of the data. This will make the noise-only output of the detection test statistic T(bold x) as constant as possible. Successful normalization makes sonar signal processing much simpler. For example, in automatic tracking, normalizing increases the probability that true tracks get initiated, and decreases the probability that false tracks get initiated. Normalizers work by estimating the background noise power and dividing T(bold x) by that estimate. The split window normalizer (SWN) is a common method of normalization. The SWN is also known as a cell-averaging constant false alarm rate (CA-CFAR) processor. The SWN normalizes each cell of the data by finding a local noise power estimate, and dividing the cell by that local estimate. The SWN finds a local noise power estimate using data in a window around the cell to be normalized. The SWN assumes the noise is stationary in the window. So, the size of the windows must be chosen so the data is approximately stationary over the window. This limits the total amount of data available to estimate any cell. The normalizer developed in this paper is based on the minimum variance spectral estimator (MVSE). The MVSE is a power spectral density (PSD) estimator that easily extends to multiple dimensions. PSD estimators estimate power as it changes over frequency. This is mathematically similar to estimating power as it changes over range or other independent variable. A normalizer based on a PSD estimator does not assume the background power is stationary. Removing the stationary assumption allows the use of larger data windows to estimate the power of each cell. In fact, we can extend the data win- ows over all the ranges, bearings, and pings to perform a three dimensional (3D) estimate of the background power. This 3D estimate uses more data to estimate the background power than the SWN. The SWN can be extended to multiple dimensions but the data windows it uses are limited by assuming the data is stationary in the window. Since the MVSE normalizer uses more data it should produce a better estimate of the background power. The SWN is compared with the new normalizer using a simple track initiation algorithm developed in this paper. Simulation results indicate the MVSE normalizer is 3 dB to 5 dB more effective than the standard SWN for active sonar normalization.