An Easy-To-Implement Face Tracker Using Particle Filtering (Part 1)

Feel free to skip this part if you are already familiar with particle filtering! Particle Filter is a type of Monte Carlo method for estimating the internal states in dynamical systems. A particle is a guess of the current state (e.g, the speed and location of a moving robot), with a weight (probability of the guess being the true state). The main idea of the particle filter is to iteratively generate a group of such "particles" to describe the probability distribution of the true current state. The higher probability a particle carries, the more likely that particle's state will appear in the final state estimate. Particles with lower weights will be filtered out.