Writer Adaptation with Style Transfer Mapping

Abstract

Adapting a writer-independent classifier toward the unique handwriting style of a particular writer has the potential to significantly increase accuracy for personalized handwriting recognition. This paper proposes a novel framework of style transfer mapping (STM) for writer adaptation. The STM is a writer-specific class-independent feature transformation which has a closed-form solution. After style transfer mapping, the data of different writers are projected onto a style-free space, where the writer-independent classifier needs no change to classify the transformed data and can achieve significantly higher accuracy. The framework of STM can be combined with different types of classifiers for supervised, unsupervised, and semi-supervised adaptation, where writer-specific data can be either labeled or unlabeled and need not cover all classes. In this paper, we combine STM with the state-of-the-art classifiers for large-category Chinese handwriting recognition: learning vector quantization (LVQ) and modified quadratic discriminant function (MQDF). Experiments on the online Chinese handwriting database CASIA-OLHWDB demonstrate that STM-based adaptation is very efficient and effective in improving classification accuracy. Semi-supervised adaptation achieves the best performance, while unsupervised adaptation is even better than supervised adaptation. On handwritten text data, semi-supervised adaptation achieves error reduction rates 31.95 and 25.00 percent by LVQ and MQDF, respectively.

Cross-subject emotion EEG signal recognition based on source microstate analysis

“…In transfer learning, we use the training set data as the source domain and the test set data as the target domain. The style transfer mapping method maps the data from the source domain to the target domain through affine mapping, reducing the distance between the source and target domains (Zhang and Liu, 2012 ). This causes the classification model to be more familiar with the target domain data, leading to better classification results.…”

An Affective Brain-Computer Interface Based on a Transfer Learning Method

“…where • F is the Frobenius norm of a matrix, • 2 is the L2-norm of a vector, and the hyperparameters β and γ govern the tradeoff between style nontransfer and overtransfer. Similar to previous studies [18], [44], we recommend setting the hyperparameters as…”

“…2) Style Transfer Mapping: The aim of STM is to map data from source points S = {s i ∈ R d |i = 1, ..., n} to target points T = {t i ∈ R d |i = 1, ..., n} via affine mapping [18], [44]. We assume the transformation from t i to s i with a confidence of f i ∈ [0, 1] and then learn the inverse transformation from s i back to t i using an affine transformation of As i +b.…”

“…3) Mapping Method: We used learning vector quantization (LVQ) to map the source points to the target points. LVQ method can learn multiple prototypes for each sample class; then, we specified the transformation of each source point toward its nearest prototype [18], [44]. To map samples with high reliability, we first removed the support vectors (SVs) after classifier training because the SVs represent the nearest sample from the decision boundaries for each pair of classifications, which are difficult to categorize.…”