As far as I know, in order to compute the maximum Lyapunov exponent, the equation generating the original data must be restarted at a very close position and, its divergence with respect to the original data must be measured. This is possible as long as an equation is used to generate the data; however in some cases –like when analyzing heart beats or exchange rates- experimental data from real life cannot be restarted.

Then my question is: Is there any way to compute the maximum Lyapunov exponent from an allegedly chaotic time series of real life experimental data which cannot be restarted?