| 摘要 |
<P>PROBLEM TO BE SOLVED: To solve such a problem that there is an issue that a conventional method for estimating a recumbent position lacks in versatility for each person, namely, when an estimation is performed using a model learned by teacher data of a subject itself, the estimation is highly precisely performed, however, when using a learning model of another person, the precision lowers. <P>SOLUTION: The device for estimating a recumbent position is used by being connected to a sensor to be worn on a person via a network and estimates the posture of the person during sleep. The device includes: a receiving unit for receiving sensing data acquired by the sensor; a feature amount calculation unit for detecting body motion periods of the person from the sensing data and calculating feature vectors of body motions in the body motion periods; a recumbent position transition probability calculating unit for calculating for each of the plurality of detected body motion periods, a transition probability that is a change from one of a plurality of predetermined recumbent positions to another one of the plurality of recumbent positions or no change, based on the feature amount; and a recumbent position column calculating unit for calculating a maximum likelihood state column based on Markov model defined as a transition probability of each of the body motion periods. <P>COPYRIGHT: (C)2013,JPO&INPIT |
