Estimating Unbiased Average Power of Digital Signal in the Presence of High-Level Impulses

Signaling data collected in the mobile radio field environment have revealed an extremely high value of impulsive noise contamination. When the sampled data contain such high-level impulses, the impulses tend to dominate the average value calculations and could result in mean values which may not be representative of the noise power which affects the system performance (signaling and voice). A statistical method has, therefore, been developed to estimate the average power which includes the noise impulses, while retaining the effects of other forms of interference. The technique is based on the knowledge that the sampled data statistically conform to a Rician distribution, in general; and degenerate to a Rayleigh distribution when the specular component vanishes. These two extremes can be used to bound the instantaneous sample values and to "filter" those impulses which do not conform to the expected behavior. Specifically, the average values are calculated by counting the number of sampled values which exceed preset thresholds and by using the count data to estimate the means in real-time processing and in post-run data processing.