Identification of time-varying biological systems from ensemble data

Area of Science:

• Systems identification
• Signal processing
• Biomechanics

Background:

• Accurate identification of time-varying systems is crucial in many scientific and engineering fields.
• Traditional methods often require a priori assumptions about system structure or input signals, limiting their applicability.
• Understanding dynamic system behavior, especially in biological systems, requires methods capable of tracking rapid changes.

Purpose of the Study:

• To describe a novel method for identifying time-varying systems.
• To demonstrate the method's ability to estimate time-varying impulse response functions without restrictive assumptions.
• To validate the method's accuracy and robustness through simulations and a biomechanical application.

Main Methods:

• The proposed method utilizes singular value decomposition (SVD) for least-squares estimation.
• It processes an ensemble of input-output realizations to identify system dynamics.
• No prior assumptions on system structure or time-variation form are needed; input signal restrictions are minimal.

Main Results:

• Simulation studies with a time-varying joint dynamics model showed accurate tracking of rapid system changes.
• The method demonstrated robustness in the presence of output noise.
• Application to dynamic ankle stiffness during isometric contraction revealed a decrease in low-frequency gain during the transient phase, unexplainable by standard models.

Conclusions:

• The developed method provides an effective and robust approach for identifying time-varying systems.
• It accurately captures rapid dynamic changes, even with noisy data and without prior system knowledge.
• The findings challenge conventional models of joint dynamics under non-stationary conditions, highlighting the method's utility in biomechanics.