Software implementation of real-time Hurst parameter estimator algorithm with filter banks

The article describes software implementation of Hurst (self-similarity) parameter estimator based on discrete wavelet transform. The main purpose is to estimate the Hurst parameter as soon as possible, by using acquired traffic samples and avoiding segmentation and waiting for data accumulation. Thus, the processing of traffic would be evenly distributed over the entire measurement time interval, instead of data accumulation and necessary performance spikes for estimation. Such traffic processing allows estimation of its parameters, in this article the main parameter considered is the Hurst parameter (or self-similarity parameter), which affects traffic processing resources requirements greatly and thus is of the interest, when estimating resources quota for traffic. The article describes proposed algorithm and provides an example of such estimation for Hurst parameter value of 0.9. This example leads to some conclusions regarding the number of scales of the discrete wavelet transform. The algorithm proposed here is independent of implementation or software platform and can be implemented for any platform supporting C programming language.