The complex and non-stationary character of multichannel biological signals like the EMG and the EEG requires dynamic methods for multidimensional processes in the spectral domain. The paper presents an adaptive algorithm of fitting bivariate time-dependent ARMA models. On this basis it is possible to calculate both univariate and bivariate spectral parameters with high temporal and spectral resolution in a parametric way. Further, the application of mapping procedures allows a topographic dynamic analysis of multidimensional non-stationary processes. The possibility of complex dynamic spectral analysis in this way is exemplarily shown for surface EMG signals of movements with fast changes and for EEG signals of finger movement.