山越 芳樹, 岡本 恭明, 篠塚 憲男

Yoshiki YAMAKOSHI, Takaaki OKAMOTO, Norio SHINOZUKA

群馬大学工学部

Gunma University, Faculty of Engineering

キーワード : Maximum entropy method, Fetal movements, Automatic classificatin.

In order to measure fetal movements quantitatively, we have been proposed a new system based on the use of the ultrasonic pulsed Doppler technique. This measurement method has a high accuracy of up to 100 microns, and fetal movements can be observed as proper displacement data by using the method. In this paper, a method which can classify fetal movements from observed displacement data is proposed. We use three special features of fetal movements for automatic classification; i) frequency of the fundamental spectral component, ii) amplitude of the displacement data and iii) continuity of the fetal movement. To derive spectral features from the displacement data, a spectrum estimation method based on the maximum entropy method is used. The filter length and the data length of the maximum entropy method are evaluated numerically. The concrete algorithm of the method is presented and some experimental results are shown.