The paper proposes a method for the processing of BTT data deriving from non-synchronous vibrations measured at constant rotor speed by a set of non-uniformly distributed sensors. The sampled data are interpreted as members of a vector space whose characteristics are determined by the signal itself and by sampling pattern. If the signal contains a single harmonic component, its frequency can be estimated through a method that has been named harmonic matching. On the contrary, when more than one harmonic component is present, due to a multi-modal response, a component separation processing is necessary. To this purpose, it is proposed a technique based on the Independent Component Analysis (ICA). This approach is limited to constant speed regime, but has the benefit of using statistical estimators that enable a strong resistance to noise. The method is illustrated using academic examples and is employed to study a flutter-like non-synchronous vibration of a bladed disk.