Ear Detection Based on Improved AdaBoost Algorithm

Ear detection is the most important step of an ear recognition system, and the detection effect of this step directly affects the performance of the whole recognition system. according to the structural characteristics of the ear itself, this paper makes improvement on the traditional AdaBoost algorithm in view of its deficiency. There are three key contributions in this paper. The first contribution is a method which can affect the emphasis point of the detector performance in order to reduce the false alarm rate, by means of changing the weight distribution of weak classifiers.The second is the introduction of a new parameter called elimination threshold ,which can improve the robustness of the detector and prevent over fitting. With the detector that we finally obtained, we test on the database of CAS-PEAL and the other two detection databases. The test results an upwards of 97% hit rate, the experimental results indicate that the ear detecting system in this paper has good detection effect. The third contribution is we designed an ear detection system of DSP and gained a good result of practical application.