Pattern Recognition on Human Skin Tissue

Abstract

Using 20 MHz ultrasound the different layers of human skin tissue can be separated. This process of pattern recognition is done by extracting features within a twostep process including an autoregressive signal model with a weighting factor for the estimation of time dependant parameters and stationary first order discrete markov chains. The statistic classification process has to be trained by a preclassified set of training data before classifying new data by a maximum likelihood algorithm. The layers epidermis. corium, subcutaneous fat and underlying structures can be separated. The results of the classification process have been successfully tested in tests of medicaments, diagnosis of osteoporosis and supervision of therapy in psoriasis patients.