人の身体的・精神的状態のモデル化

少子高齢化,医療費高騰の問題に限らず,我々の身体的・精神的な健康を維持・回復することは自立した人間として生きていく上でもっとも重要なことである. そのためには,人の身体的・精神的な状態を,常に生活に不便・悪影響の無い方法で計測し,そのモデル化・解析結果に基づいて,健康・正常な状態を維持・回復するための方法を人に提案していくための研究を進める. そのための広範囲に渡る研究テーマを,特に大量・多人数・長期間のデータからの人のモデル化(身体モデル,動きモデル,精神モデル,脳モデル,など)を取り扱う. 主に,

などを研究対象とする.

過去・現在の具体的な研究テーマを以下に示す.


・身体動作修正のための人体計測・可視化

Our goal is to develop a system for coaching human motions (e.g. rehabilitation). Common motion measurement systems are too expensive and require users to wear binding devices. The proposed system utilizes an inexpensive depth-measurement sensor (i.e. Microsoft Kinect). in order to get high-measurement accuracy with no body-equipped devices. The system functionally consists of three modules below. The first one estimates the sequence of a body pose from a depth image sequence captured while a user performs a target motion. The second one evaluates the gaps between the sequences of the estimated poses and a good template. The third one coaches the user on how to modify his/her motion so that it gets closer to the good template. This work focuses on achieving the third point. To this end, it is important to efficiently advise the user to emulate the crucial features that define the good template. This is because many other features of the target motion might be varied among individuals, but those variations give less impacts on evaluating the target motion. The proposed method automatically mines the crucial features of any kind of motions from a set of all motion features. The crucial features are mined based on feature sparsification through binary classification between the samples of good and other motions. Experimental results demonstrated that 1) the proposed method could extract intuitively-correct crucial features (as shown in Fig. 1) and 2) the extracted features improved the accuracy in classifying good and other motions.

図1: 抽出された動き修正のための重要な要素

・身体動作修正のためのインタフェース

The development of a widely applicable automatic motion coaching system requires to address a lot of issues including motion capturing, motion analysis and comparison, error detection as well as error feedback. In order to cope with this complexity, most existing approaches focus on a specific motion sequence or exercise. As a first step for the development of a more generic system, this work systematically analyzes different error and feedback types. A prototype of a feedback system that addresses multiple modalities is presented. The system allows to evaluate the applicability of the proposed feedback techniques for arbitrary types of motions in a next step. The screenshot of the developed interface system is shown in Figure 2.

図2: インタフェースのスクリーンショット

・Human Body-parts Tracking for Fine-grained Behavior Classification

This work discusses the usefulness of human body-parts tracking for acquiring subtle cues in social interactions. While many kinds of body-parts tracking algorithms have been proposed, we focus on particle filtering-based tracking using prior models, which have several advantages for researches on social interactions. As a first step for extracting subtle cues from videos of social interaction behaviors, the advantages, disadvantages, and prospective properties of the body-parts tracking using prior models are summarized with actual results.

図3: 類似動作の低次元モデル化(上が歩行で,下がジョギング)