A human fall detection method and protection device
A detection method and protection device technology, applied to instruments, alarms, etc., can solve the problem that the fall protection device is difficult to function, achieve good protection and avoid accidental injuries
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment
[0034] Example: see figure 1 As shown, this embodiment provides a human fall detection method, including: acquiring training data, performing feature extraction, obtaining training samples, training a classification model, and obtaining a trained classifier. In order to test its effect, the test data is obtained, feature extraction is carried out, and the trained classifier is used for classification decision-making.
[0035] Among them, when performing feature extraction, the specific considerations for selecting feature values are expressed as follows:
[0036] The sampling frequency in this embodiment is 100Hz.
[0037] (1) Selection of acceleration features:
[0038] In the process of human body falling, the acceleration will change obviously. Considering the different falling directions of front, back, left, and right, the three-axis combined acceleration is taken as:
[0039] ;
[0040] figure 2 Shown is the resultant acceleration curve of a forward fall. It ca...
Embodiment 2
[0089] Embodiment 2: Using the method described in Embodiment 1, the length of the sliding window is changed to conduct a training test.
[0090] See attached Figure 4 As shown, it is the acceleration vector and change curve of a walking process. First, define the window length w and the superposition length o of the window. For a period of time series data , the first window is denoted as , extract acceleration and angular velocity features from the data in this window. Since the stacking length of the window is o, the data of the next window is expressed as , continue to process the data in this window, and so on, so one daily behavior can collect multiple sets of daily behavior samples. The completion time of each daily behavior in the experiment is 5 seconds, the sliding window length w of daily behavior sample collection is also set to 100ms, and the window superposition rate is set to 50%, that is, the last 50% data of the current window is used as the first 50% d...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More