We are studying to estimate the behavior of care recipients and patients in bed from the sounds they generate. Since the distribution of the sound source and its movement, and spectral information are supposed to be the input information for the behavior estimation, it is important to generate high reliable sound source image against friction sound which is the most of action sound. By installing microphone arrays in units of microphone pairs with relatively small spacing on the headboard and footboard, and comparing the amplitude strength of the cross-correlation and delay sum for each microphone pair with the diagonal microphone pair, I construct an estimation method stable for friction action sound and improving longitudinal resolution.
J-STAGEがリニューアルされました! https://www.jstage.jst.go.jp/browse/-char/ja/