基于HHT的脱机手写体汉字特征提取

Feature extraction of off-line handwritten Chinese characters based on Hilbert-Huang transform

提出了一种基于数学形态学方向笔画提取的希尔伯特—黄变换(HHT)方法,对脱机手写体汉字进行特征提取。通过对HHT得到的Hilbert谱和边际谱的分析,构成4维结构特征和32维统计特征,并对其进行特征融合后得到36维特征向量作为最终的识别特征。试验结果表明其识别率比单独使用Gabor变换、小波变换等方法的识别率高。在识别速度上虽然比矩变换、数学形态学等方法慢,但是比Gabor变换的速度有明显提高,比多方法特征融合的方法在速度上有一定提高。该研究表明HHT作为一种新的信号分析方法,可以被有效地运用于提取汉字图像的特征。

Based on mathematical morphology, the authors studied a new feature extraction approach for off-line handwritten Chinese characters by Hilbert-Huang Transform （HHT）. After an analysis of Hilbert spectrum and marginal spectrum extracted by HHT, 4-dimensional structure features and 32-dimensional statistical features were reached. A 36- dimensional vector gained by the feature fusion was the final recognition feature. Experimental results show that this method has a higher recognition rate than Gabor transform, wavelet transform etc., and it also has a higher recognition speed than most of the feature extraction methods available. It indicates that HHT can be effectively applied to the feature extraction of Chinese characters.