LAUSR

Abstract The article provides a proposal for the processing of a time-frequency plane, obtained by STFT and spectrogram, to extract the characteristics features of VAG vibroarthrographic signals of the knee joint. The authors present three new descriptors, enabling initial diagnosis in the direction of classification to the following disease classes: healthy knee, I- II- III- stage of chondromalation and osteoarthritis. Descriptors are based on the time-mariginal condition, obtained by cutting a portion of the time-frequency plane of the spectrogram and its low-pass filtration. Descriptors consider following features of the time-marginal condition: averaging of local maximum, number of maximum and averaging values of slopes. Research were carried out on 184 patients. Research prove significant diagnostic potential of the descriptors. Moreover white noise influence on the results were also investigated.