We describe a new method of overcomingproblems inherent in peculiar velocity surveys by using data compression as a filter with which to separate large-scale, linear flows from small-scale noise that biases theresults systematically. We demonstrate the effectiveness of our method using realisticcatalogs of galaxy velocities drawn from N--body simulations. Our testsshow that a likelihood analysis of simulated catalogs that uses all ofthe information contained in the peculiar velocities results in a biasin the estimation of the power spectrum shape parameter $\Gamma$ andamplitude $\beta$, and that our method of analysis effectively removesthis bias. We expect that this new method will cause peculiar velocitysurveys to re--emerge as a useful tool to determine cosmologicalparameters.
