A Human Pose Discrimination Method Based on Support Vector Machine

Abstract

The present invention proposes a human body attitude discrimination method based on a support vector machine, comprising the steps of: using a sliding window mechanism to segment the sample data, dividing the original signal into several subsets; calculating the mean value of the triaxial combined acceleration of each subset, The standard deviation, covariance and skewness coefficient values ​​of the three-axis combined acceleration are used as the eigenvectors of the corresponding subsets, and the classification label values ​​are set for each subset; each corresponding fuzzy factor is calculated according to the eigenvectors; using the fuzzy factors and The classification label value trains the support vector machine to obtain the optimal hyperplane decision function for attitude classification; the trained support vector machine is used as a classifier to classify the resampled human triaxial acceleration data. The present invention adopts a data-oriented machine learning method to discriminate the fall posture, and replaces the traditional method of using the binorm based on the triaxial acceleration value of the human body as a threshold to judge whether a person has fallen, thereby improving the detection rate.

Description

technical field [0001] The invention relates to the field of human body posture discrimination, in particular to a human body posture discrimination method based on a support vector machine. Background technique [0002] The gradual aging of the population around the world today has made the entire international community pay more and more attention to the issue of aging. Falling is a frequent accident of the elderly, which seriously affects the physical and mental health, and even threatens the safety of life. According to the report of the National Safety Council of the United States, among the population over the age of 65, the death caused by falls ranks first among all accidental deaths, accounting for 33% of accidental deaths in this age group. In the traditional method for human body posture recognition, the two-norm based on the three-axis acceleration value of the human body is generally used as the threshold to judge whether the person has fallen. This method has ...

AI Technical Summary

Problems solved by technology

In the traditional method for human posture recognition, the two-norm based on the three-axis acceleration value of the human body is generally used as the threshold to judge whether the person has fallen. This method has poor measurement accuracy and is prone to false alarms.

Images