Co-training approach for improving age range prediction from handwritten text

Recently, the combination of classification systems with semi-supervised learning has attracted researchers in several fields. Usually, for tasks with high complexity such as handwriting based age prediction, individual systems, using one classifier associated with specific data features, cannot provide satisfactory performance. In this paper, we investigate the contribution of the Co-training approach, as an enhancement procedure, for age range prediction from handwriting analysis. By using samples extracted from IAM dataset, several descriptors for feature generation and an SVM predictor for classification are employed. The comparison with the state of the art indicates the effectiveness of this approach.