An adaptive classifier fusion method for analysis of knee-joint vibroarthrographic signals

Externally recorded knee-joint vibroarthrographic (VAG) signals bear diagnostic information related to degenerative conditions of cartilage disorders in a knee. In this paper, the number of atoms derived from wavelet matching pursuit (MP) decomposition and the parameter of turns count with the fixed threshold that characterizes the waveform variability of VAG signals were extracted for computer-aided analysis. A novel multiple classifier system (MCS) based on the adaptive weighted fusion (AWF) method is proposed for the classification of VAG signals. The experimental results shows that the proposed AWF-based MCS is able to provide the classification accuracy of 80.9%, and the area of 0.8674 under the receiver operating characteristic curve over the data set of 89 VAG signals. Such results are superior to those obtained with best component classifier in the form of least-squares support vector machine, and the popular Bagging ensemble method.