| 摘要 |
Provided are a method and a device for recognizing a motion by collecting motion measurement data from a sensor for measuring a motion of a human body such as an acceleration sensor, a gyro sensor, and the like and statistically analyzing the collected data. The method for recognizing a motion of the device for recognizing a motion comprises a step of receiving motion measurement data from a sensor attached on a recognition target; a step of computing a tilt angle sum (TAS), sum of the variance of accelerations (SVA), and autocorrelation coefficient of tilt angle (ACT) from the motion measurement data; a step of modeling the distribution of data points composed of the TAS, the SVA, and the ACT using a Gaussian mixed model; and a step of non-parametric Bayesian clustering the modeled data points by estimating aposteriori distribution with respect to infinite Gaussian mixed model using collapsed Gibbs sampling. |
