ウェーブレット変換を適用した深度情報による離床判定システムの構築

抄録

In this study, we proposed an image processing method to detecting bed-leaving behavior. In our proposed method, feature vector is obtained by applying the wavelet transform to depth images that provide distance information between the camera and the target. By classifying the obtained feature vectors using Self-Organizing Map, our proposed method can classify three postures: lying down, turning over, and leaving the bed. We applied the proposed method to three targets and were able to classify three postures: lying down, turning over, and leaving the bed. The proposed method was also effective for unlearned data.