Classifiers under Continuous Observations

Many researchers have reported that recognition accuracy improves when several images are continuously input into a recognition system. We call this recognition scheme a continuous observation-based scheme (CObS). The CObS is not only a useful and robust object recognition technique, it also offers a new direction in statistical pattern classification research. The main problem in statistical pattern recognition for the CObS is how to define the measure of similarity between two distributions. In this paper, we introduce some classifiers for use with continuous observations. We also experimentally demonstrate the effectiveness of continuous observation by comparing various classifiers.