Kalman Filtering of Noisy Video Tracking Data from Morris Water Maze Tests

The Kalman filter algorithm is applied to provide estimates of position and velocity from noisy position measurement data. Position data were obtained from a vision system designed to track the swimming path of a rat to measure spatial memory and recognition functions in a setup called the Morris water maze test. These tests are used with rats to study neurological functions. Although the computer system is reliable, the data can be disturbed, either by effects of changes in lighting, or mismatch in camera placement. In this work, it is shown that the Kalman algorithm has satisfactory performance in the elimination of large noise data values, and that it is very important to consider the measurement noise as a way to force the algorithm to underestimate such measurements, which contain remarkable errors. The results show the great ability of the algorithm to model the noise and reduce its effect. However, there is a compromise between noise reduction level and excessive data smoothing. Numerical values for distance, average velocity and total time for several tests are given. This approach can be applied to analyze stored data as well as to real-time data.