Enhancing Age Estimation from Handwriting: A Deep Learning Approach with Attention Mechanisms

Abstract

Currently, age estimation is a hot research topic in the field of forensic biology. Age estimation methods based on facial or brain features are easily affected by external factors. In contrast, handwriting analysis is a more reliable method for age estimation. This paper aims to improve the accuracy and efficiency of age prediction using handwriting analysis by proposing a novel method that integrates a coordinate attention mechanism in a deep residual network (CA-ResNet). This method can more accurately capture important features in the input handwritten images while reducing the number of model parameters, thereby improving the accuracy (Acc) and efficiency of the model for age estimation. The proposed method is evaluated on standard handwriting datasets and the created dataset, and it is compared with the current state-of-the-art methods. The results show that the method consistently outperforms others, achieving an accuracy of 79.60% on the IAM handwriting dataset, with a 6.31% improvement over other methods.