In-air handwritten Chinese character recognition using multi-stage classifier based on adaptive discriminative locality alignment

The in-air handwriting is a natural and useful way for human-computer interaction. Yet, to our knowledge, few work has been done for the in-air handwritten Chinese character recognition (IAHCCR). In this paper, we present a multi-stage recognizer to address the problem of IAHCCR. The proposed methods can also deal with the classical handwritten Chinese character recognition (HCCR). We find that the discriminative locality alignment (DLA) technique in HCCR heavily depends on the choice of parameters in practice. To overcome the disadvantage, we present an adaptive discriminative locality alignment (ADLA), which does not involve the parameter optimization process. At the same time, a new static similar characters collection technique is proposed. We evaluate the proposed methods on the IAHCC-UCAS2014 dataset, an in-air handwritten Chinese character dataset constructed by us, as well as the SCUT-COUCH2009 database, a HCCR dataset. The experimental results demonstrate the effectiveness of the proposed methods on two different kinds of dataset.

1. Introduction

In the past 40 years, many impressive achievements in handwriting Chinese character recognition (HCCR) have been made. However, the recognition of unconstrained handwritten characters is still a challenge [1], [2]. As human-computer interaction devices become diversified, new application requirements arise in HCCR, for example, the fingertip in-air writing system [3] in recent years. The 3D in-air writing is a natural and useful way for human-computer interaction. Different from touch-screen handwriting, the in-air written character has no pen-lift information, i.e., a character is always finished writing in one stroke. The in-air writing with fingertip is not easy to control to produce smooth and formal trajectories, so the structure of generated characters is more complex and varied in intra-class, and the jitter of stroke is more evident. Although many non-strokes are used to linked different real strokes, they are also recorded by the sensors as strokes so that a character consists of only one stroke, which causes much more similar characters than touch-screen writing. Fig. 1 gives some examples to intuitively show the difference between touch-screen handwritten characters [4], [5] and in-air written characters. Fig. 1.

Examples of similar characters for chinese