Effect of discrete cosine and wavelet transformation based compression on the long range dependence of communication network performance measurements

2007-10-26T08:09:51Z (GMT) by Kostas Kyriakopoulos David Parish
This paper examines the impact of compression methods on the long-range dependence of communication network traffic measurements. The two compression methods that are examined are based on the Wavelet transformation and the Discrete Cosine Transformation (DCT). In order to measure the length of long-range dependence of a stochastic process, we first have to estimate the Hurst parameter. The Hurst parameter is estimated by using the rescaled range statistic (R/S) method. The Hurst values of the examined signal, before and after the applied compression, are estimated and compared. If the Hurst value of the compressed signal is close to the Hurst value of the uncompressed signal, then the compression algorithm has little interference on the longrange dependence. The results show that Wavelet transformation performs better than the DCT.