This paper investigates methods for analysing dynamics of biological systems characterised by multimodal signals of different dimensionality, such as polysomnography (sleep) signals or population tracking. We study statistical properties of low dimensional representations of multimodal signals, which we want to use as features for detection of intrinsic signal state changes. We assume that distinctive state transition features remain preserved in low dimensional signal representations of dynamical system signals. We first build recurrence plots for well-behaved biological systems (predator-prey model and Lorenz system which models a forward osmosys). Then, we provide recurrence plot analysis for the real dataset (visual stimulation of ferrets) and we show that preliminary statistical analysis of ferret datasets has stable characteristics. This motivates us to further pursue study on sleep signal statistics that we plan to use for sleep stage detection.