The performance of real time applications such as video andvoice streams relies on packet delay jitter. Large delay jittercauses buffer overflow or underflow at the receiver end and theuser encounters interrupts. The delay jitter is mainly due to theperturbation of background traffic in the bottleneck router.Fulton and Li provide an analytical approximationfor the first-order and second-order statistics of delay jitter.However, their analysis is based on a Markovian model of thebackground traffic, which is not quite suitable for Internettraffic and requires lots of computational effort. We propose anefficient method to predict the jitter variance of a CBR(constant-bit rate) connection based on the wavelet model of thebackground traffic. The wavelet analysis extracts the statisticalproperties of background traffic and the analysis result can beused to predict an upper bound for the jitter variance of the CBRconnection.