diff --git a/blog/assets/images/biometrics-explained/kinect-skeleton.jpeg b/blog/assets/images/biometrics-explained/kinect-skeleton.jpeg new file mode 100644 index 00000000..79fced28 Binary files /dev/null and b/blog/assets/images/biometrics-explained/kinect-skeleton.jpeg differ diff --git a/blog/posts/biometrics-explained.md b/blog/posts/biometrics-explained.md index 4292c328..64681039 100644 --- a/blog/posts/biometrics-explained.md +++ b/blog/posts/biometrics-explained.md @@ -144,6 +144,11 @@ There are two approaches for extracting biometric data from video. This approach attempts to model the human body in order to track the different parts of it. A well-known example of this approach is the Microsoft Kinect, which only consists of a fairly low-resolution camera. It simplifies the human body into a stick figure, which you can see in footage of the [Kinect](https://www.youtube.com/watch?v=33AsuE-WP64) in action. It then uses the distances and joint angles of the model for gait recognition. +
+ ![Diagram showing the kinect's model of the human body, a 2D skeleton made up of various parts](../assets/images/biometrics-explained/kinect-skeleton.jpeg) +
source: Alireza Bilesan, Saeed Behzadipour, Teppei Tsujita, Shunsuke Komizunai, and Atsushi Konno
+
+ #### Model-free Model-free approaches don't try to model the human body but instead use the whole motion of human silhouettes. This gives a few advantages, namely it works regardless of camera quality and it takes significantly fewer resources.