Cleaning method based on iterative and interactive perceived age database

The invention discloses a cleaning method based on an iterative and interactive perceived age database. The method comprises the following steps that: firstly, training a physiological age database byan SVM (Support Vector Machine) to obtain a classifier A, and then, utilizing the classifier A to identify the perceived age database; adding a sample number which is correctly identified into the physiological age database to form a new training set, training the new training set by the SVM to obtain a classifier B, and utilizing the classifier B to identify the wrongly identified samples of theclassifier; repeatedly circulating, stopping iteration when the fluctuation range of age identification accuracy is within 0.1%, adopting a human-computer interaction way to correct the age label ofthe sample for the left wrongly identified samples after iteration stops, and repeating the above steps for the corrected samples; and stopping circulation until all samples are wrongly identified, and finally, subtracting the