transformBasis2.m this is the same as transformBasis.m, but it inverts the covariance matrix after taking the cholesky factorization. sometimes this seems to be numericaly stable whereas transformBasis.m isnt. transforms the original waveforms onto an other basis,defined by the autocorrelation matrix, such that datapoints are uncorrelated. covar is the covariance matrix see pouzat et al 2003, J Neurosci methods see Kay, vol1, pp 94-95 for pre-whitening urut/nov04