Adaptive spectral analysis of non-stationary vibroarthrographic signals using autoregressive modelling

Vibroarthrographic (VAG) signals are obtained during dynamic knee movements and hence are essentially non-stationary. Spectral analysis (the spectrum at a given joint angle or a given time for a constant joint angular velocity) of VAG signals can serve as a valuable tool for diagnosis of cartilage pathology in the joint. This paper presents the application of adaptive autoregressive modeling to VAG signal spectral analysis. Results indicate that this method can provide frequency and time (joint angle) information of the signal along with good frequency-time-localization at different knee joint angles. The adaptive AR spectral analysis technique provides better performance as compared to conventional linear prediction and Fourier methods.