Chinese Character Identification Research Articles

A Novel Multi-Stage Recognition System in Handwritten Chinese Characters

Due to the various individual handwriting styles and diverse structures of Chinese characters, it is rather intricate to recognize handwritten Chinese characters precisely. In this paper, a muti-stage system that combines a Differentiable Binarization Network (DBNet) for text detection and Scene Text Recognition with a Single Visual Model (SVTR) for effective character recognition is developed, aiming to offer a potential solution for correct identification of handwritten Chinese characters in real-world scenarios. More specifically, the proposed system utilized the CASIA-HWDB2.x dataset from the Institute of Automation of the Chinese Academy of Sciences to train text detection and recognition models. For the detection, the Differentiable Binarization method in the Head Network applies a dynamic threshold map for text-background differentiation. On top of the SVTR employed as a recognition module, this research proposed a performance-optimized PP-LCNet architecture and the Guided Training of Continuous Time-Series Classifiers (GTC) with TextConAug, which is a data augmentation strategy to accommodate real-world scenarios. Briefly, DBNet and SVTR are integrated to form an end-to-end system. The output of DBNet after undergoing cropping and direction classifier, guides SVTR for a comprehensive system, ensuring seamless detection and recognition of handwritten Chinese characters. Experimental results indicate that this system achieves a promising performance and can provide accurate information on handwritten Chinese characters.

Evaluation and Recognition of Handwritten Chinese Characters Based on Similarities

To accurately recognize ordinary handwritten Chinese characters, it is necessary to recognize the normative level of these characters. This study proposes methods to quantitatively evaluate and recognize these characters based on their similarities. Three different types of similarities, including correlation coefficient, pixel coincidence degree, and cosine similarity, are calculated between handwritten and printed Song typeface Chinese characters. Eight features are derived from the similarities and used to verify the evaluation performance and an artificial neural network is used to recognize the character content. The results demonstrate that our proposed methods deliver satisfactory evaluation effectiveness and recognition accuracy (up to 98%~100%). This indicates that it is possible to improve the accuracy in recognition of ordinary handwritten Chinese characters by evaluating the normative level of these characters and standardizing writing actions in advance. Our study can offer some enlightenment for developing methods for the identification of handwritten Chinese characters used in transaction processing activities.

Development and validation of a Chinese pseudo-character/non-character producing system.

This study developed and validated a Chinese pseudo-character/non-character producing system (CPN system) that can assist researchers in creating experimental materials using Chinese characters. Based on a large-scale dataset of 6097 characters, the CPN system provides researchers with precise Chinese orthographic information (structures and positions, radical frequency, number of strokes, number of radical-sharing neighbors, and position-based regularity) to create three types of experimental stimuli: pseudo-characters, semi non-characters, and whole non-characters. Featuring the position-based regularity of 446 radicals, the CPN system helps researchers to manipulate, or to control for, orthographic characteristics of radicals to study Chinese lexical processing. In two empirical validations for stimuli created by the system, Chinese-as-second-language learners (n = 79) and first-language users (n = 41), respectively, participated in a Chinese orthographic choice task in which participants compared two artificial characters and chose the one that more closely resembled a real Chinese character. Both validations demonstrate that highly proficient Chinese readers are better able to identify pseudo-characters, suggesting that the radical's position-based information impacts Chinese character identification to different extents. With the empirical support for the created stimuli, the system further affords researchers auto-generated outcomes with downloadable images and Excel sheets for creating customized stimuli, making material selection easy, efficient, and effective. This CPN system is the first large-scale, data-driven tool free for researchers who are interested in studies of written Chinese. CPN should benefit the field of Chinese orthographic processing, Chinese instruction, and cross-linguistic comparisons, providing a useful tool for studying Chinese lexical processing.

MyOcrTool: Visualization System for Generating Associative Images of Chinese Characters in Smart Devices

Majority of Chinese characters are pictographic characters with strong associative ability and when a character appears for Chinese readers, they usually associate with the objects, or actions related to the character immediately. Having this background, we propose a system to visualize the simplified Chinese characters, so that developing any skills of either reading or writing Chinese characters is not necessary. Considering the extensive use and application of mobile devices, automatic identification of Chinese characters and display of associative images are made possible in smart devices to facilitate quick overview of a Chinese text. This work is of practical significance considering the research and development of real-time Chinese text recognition, display of associative images and for such users who would like to visualize the text with only images. The proposed Chinese character recognition system and visualization tool is named as MyOcrTool and developed for Android platform. The application recognizes the Chinese characters through OCR engine, and uses the internal voice playback interface to realize the audio functions and display the visual images of Chinese characters in real-time.

Cross-Linguistic Word Recognition Development Among Chinese Children: A Multilevel Linear Mixed-Effects Modeling Approach.

The effects of psycholinguistic variables on reading development are critical to the evaluation of theories about the reading system. Although we know that the development of reading depends on both individual differences (endogenous) and item-level effects (exogenous), developmental research has focused mostly on average-level performance, ignoring individual differences. We investigated how the development of word recognition in Chinese children in both Chinese and English is affected by (a) item-level, exogenous effects (word frequency, radical consistency, and curricular grade level); (b) subject-level, endogenous individual differences (orthographic awareness and phonological awareness); and (c) their interactive effect. We tested native Chinese (Putonghua)-speaking children (n = 763) in grades 1 to 6 with both Chinese character and English word identification (lexical) decision tasks. Our findings show that (a) there were effects of both word frequency and age of acquisition in both Chinese and English, but these item-level effects generally weakened with increasing age; (b) individual differences in phonological and orthographic awareness each contributed to successful performance; and (c) in Chinese, item-level effects were weaker for more proficient readers. We contend that our findings can be explained by theoretical models that incorporate cumulative learning as the basis for development of item-level effects in the reading system.

Multi-Font Arabic Isolated Character Recognition Using Combining Machine Learning Classifiers

Nowadays, optical character recognition is one of the most successful automatic pattern recognition applications. Many works have been done regarding the identification of Latin and Chinese characters. However, the reason for having few investigations for the recognition of Arabic characters is the complexity and difficulty of Arabic characters identification compared to the others. In the current work, we investigate combining multiple machine learning algorithms for multi-font Arabic isolated characters recognition, where imperfect and dimensionally variable input charactersare faced. To the best of our knowledge, there is no such work yet available in this regard. Experimental results show that combined multiple classifiers can outperform each individual classifier produces by itself. The current findings are encouraging and opens the door for further research tasks in this direction.

Applied Research on “Fallibility” Chinese Character Aesthetics in Package Design

Chinese character is a symbol with the function of information communication, its design thought and method are of widespread significance to package design. The thesis has introduced the “Fallibility” ideas into Chinese character design to analyze the relation between “Fallibility” of Chinese characters and the identification of Chinese characters, and how to shape “Fallibility” features. Visual communication needs to break people’s common visual habits to manifest most abundant and profound content in simplified and symbolized form, besides, by extracting highly precise graphic language and easily understandable arrangement order, it can deliver expected connotation. The thesis studies how to use “Fallibility” idea for reference to form directive design method, so as to make Chinese character more flexible and exploit more free space for theoretical research and practice of package design.

Music-reading expertise modulates the visual span for English letters but not Chinese characters.

Recent research has suggested that the visual span in stimulus identification can be enlarged through perceptual learning. Since both English and music reading involve left-to-right sequential symbol processing, music-reading experience may enhance symbol identification through perceptual learning particularly in the right visual field (RVF). In contrast, as Chinese can be read in all directions, and components of Chinese characters do not consistently form a left-right structure, this hypothesized RVF enhancement effect may be limited in Chinese character identification. To test these hypotheses, here we recruited musicians and nonmusicians who read Chinese as their first language (L1) and English as their second language (L2) to identify music notes, English letters, Chinese characters, and novel symbols (Tibetan letters) presented at different eccentricities and visual field locations on the screen while maintaining central fixation. We found that in English letter identification, significantly more musicians achieved above-chance performance in the center-RVF locations than nonmusicians. This effect was not observed in Chinese character or novel symbol identification. We also found that in music note identification, musicians outperformed nonmusicians in accuracy in the center-RVF condition, consistent with the RVF enhancement effect in the visual span observed in English-letter identification. These results suggest that the modulation of music-reading experience on the visual span for stimulus identification depends on the similarities in the perceptual processes involved.

Visual Antipriming Effect: Evidence from Chinese Character Identification.

Marsolek et al. (2006) have differentiated antipriming effects from priming effects, by adopting a novel priming paradigm comprised of four phases that include a baseline measurement. The general concept of antipriming supports the overlapping representation theory of knowledge. This study extended examination of the Marsolek et al. (2006) paradigm by investigating antipriming and priming effects in a series of Chinese character identification tasks. Results showed that identification accuracy of old characters was significantly higher than baseline measurements (i.e., the priming effect), while identification accuracy of novel characters was significantly lower than baseline measurements (i.e., the antipriming effect). This study demonstrates for the first time the effect of visual antipriming in Chinese character identification. It further provides new evidence for the overlapping representation theory of knowledge, and supports generalizability of the phenomenon to Chinese characters.

N170 Changes Show Identifiable Chinese Characters Compete Primarily with Faces Rather than Houses.

Character processing is a crucial cognitive skill that is highly emphasized and industriously cultivated in contemporary society. In the present study, using a competition paradigm, we examined the electrophysiological correlates of different relationships between Chinese characters and faces and between Chinese characters and houses during early visual processing. We observed that identifiable Chinese characters compete primarily with faces rather than houses at an early visual processing stage, with a significantly reduced N170 for faces but not for houses, when they were viewed concurrently with identifiable characters relative to when they were viewed concurrently with unidentifiable characters. Consistent with our previous study, there was a significant increase in N170 after characters have been learned, indicating a modulatory effect of Chinese character identification level on N170 amplitude. Furthermore, we found an enlarged N170 in response to faces compared to houses, indicating that the neural mechanisms for processing faces and houses are different at an early visual processing stage.

Effect of Earthquake Experience on the Perceptual Interference with Unfilled Intervals in Chinese Characters

The perceptual interference effect states that a moderately degraded stimulus is less readily identified when depicted in a sequence of exposures from a highly degraded form than when presented all at once (standard condition). Thirty-two middle-school students from Mianzhu City (an earthquake-prone area), Sichuan province, China, were selected as major (experimental) participants, while thirty-two middle-school students from Beibei, Chongqing were chosen as control participants. This study, which consisted of two experiments (major and control), explored the effect of earthquake experience on the identification of Chinese characters with a paradigm of perceptual interference with unfilled interval (blank screen between exposures). The participants in each experiment were required to identify fragmented target characters in eight conditions by typing down the identity of each target. These conditions resulted from a within-participant design of 2 (material conditions: earthquake-related and unrelated characters) × 4 (presentation conditions, namely, three incremental conditions with interval durations of 0, 450, 900 ms, and the standard condition). These two experiments were combined into a dummy experiment of a mixed design for further analyses. Repeated measures analyses of variance were first conducted on percentages of the correct completion of target fragments. The major experiment indicated a significant difference in the performance between material conditions. The interference effect remained for earthquake-related characters in the critical 450 ms condition, while the effect for the unrelated characters disappeared. By contrast, the control experiment showed no difference in performance between material conditions. More critically, the results found no difference in the identification of the unrelated characters (i.e., recognition ability) between the control and major participants, yet the interference effect disappeared even for the earthquake-related characters in the 450 ms condition among the control participants. Based on these results, the identification performance of the participants with earthquake experience was higher on earthquake-related characters than the unrelated characters, and the perceptual interference for normal (unrelated) characters disappeared on the duration of 450 ms. However, the interference for the earthquake-related characters remained. The notion of persistent activation of representation cross unfilled intervals in the competitive activation model could explain why sometimes the interference occurred at the interval of 450 ms, and why sometimes it disappeared. The present study also supported the view that emotion (such as fear) drives perceptual processing.

Using singular value decomposition to investigate degraded Chinese character recognition: evidence from eye movements during reading.

Previous research indicates that removing initial strokes from Chinese characters makes them harder to read than removing final or internal ones. In the present study, we examined the contribution of important components to character configuration via singular value decomposition. The results indicated that when the least important segments, which did not seriously alter the configuration (contour) of the character, were deleted, subjects read as fast as when no segments were deleted. When the most important segments, which are located in the left side of a character and written first, were deleted, reading speed was greatly slowed. These results suggest that singular value decomposition, which has no information about stroke writing order, can identify the most important strokes for Chinese character identification. Furthermore, they also suggest that contour may be correlated with stroke writing order.

动词性义符与名词性义符在汉字识别中作用的比较<br>The Comparison of the Role of Verbal and Nounal Semantic Radicals in the Identification of Chinese Cha-racters

采用2(词类：动词、名词) × 4(启动方式：R+S+、R+S–、R–S+、R–S–)的重复测量设计，考察在启动字和目标字有相同的义符和语义(R+S+)、启动字和目标字有相同的义符和不同的语义(R+S–)、启动字和目标字有不同的义符和相同的语义(R–S+)、启动字和目标字有不同的义符和语义(R–S–)的启动条件下，对作为目标字的动词和名词的判别情况。结果显示，义符具有加快和阻碍汉字识别的双重作用，动词的识别要优于名词的识别，动词性义符比名词性义符在汉字识别中起的作用更大。 The present study used a repeated measures design of 2 (target type: verbs, nouns) × 4 (prime type: R+S+, R+S–, R–S+, R–S–) to investigate the role of verbal and nounal semantic radicals in the identification of Chinese characters. The R+S+, R+S–, R–S+, and R–S– conditions indicated that the target characters were preceded by primes that 1) shared the targets’ radicals and were semantically related (R+S+); 2) shared the targets’ radicals but were not semantically related (R+S–); 3) did not shared the targets’ radicals but were semantically related (R–S+); and 4) did not share the targets’ radicals and were not semantically related (R–S–). The experiment results showed that 1) semantic radicals had a dual function in facilitating or inhibiting character recognition; 2) verbal characters had stronger effects in facilitating character recognition than nouns; 3) verbal semantic radicals had a stronger facilitating effect than non-verbal radicals on character decision task.

动词性义符与名词性义符在汉字识别中作用的比较<br>The Comparison of the Role of Verbal and Nounal Semantic Radicals in the Identification of Chinese Cha-racters

动词性义符与名词性义符在汉字识别中作用的比较<br>The Comparison of the Role of Verbal and Nounal Semantic Radicals in the Identification of Chinese Cha-racters

Crowding of Upright Chinese Character is Stronger with Inverted than Upright Flankers: An Exception of the Similarity Rule

Upright flanking faces have stronger detrimental effects on the recognition of upright target face, than inverted flanking faces. One possible explanation for this “flanker- inversion effect” was that the more holistically processed upright flanking faces allowed for more erroneous feature integration. Alternatively, crowding was known to be stronger when target and flankers were more similar. Here we investigate flanker-inversion effect on crowding in Chinese character identification. Five normally-sighted young adults participated. Targets of size 1.2° were presented at 5° in the lower visual field. Four flankers with center-to-center distance of 1.8° were presented in the crowded condition. Three types of flankers were used, upright or inverted Chinese and upright Korean characters. The identification contrast thresholds were estimated by QUEST and crowding strength was measured through threshold elevation (TE). Crowding on upright Chinese target was significantly stronger with inverted Chinese flankers (TE = 1.59±0.32) than with upright Chinese flankers (TE = 1.47±0.29). No inversion effect was observed for inverted Chinese target. Korean flankers produced similar crowding as upright Chinese flankers. Our results go against the similarity rule that predicts upright Chinese flankers would produce stronger crowding for upright Chinese target. Holistic processing preferred for inverted Chinese characters may account for the findings.

Using stroke removal to investigate Chinese character identification during reading: evidence from eye movements

We explored the effect of stroke removal from Chinese characters on eye movements during reading to examine the role of stroke encoding in character identification. Experimental sentences were comprised of characters with different proportions of strokes removed (15, 30, and 50%), and different types of strokes removed (beginning, ending, and strokes that ensured the configuration of the character was retained). Reading times, number of fixations and regression measures all showed that Chinese characters with 15% of strokes removed were as easy to read as Chinese characters without any strokes removed. However, when 30%, or more of a character’s strokes were removed, reading characters with their configuration retained was easiest, characters with ending strokes removed were more difficult, whilst characters with beginning strokes removed were most difficult to read. The results strongly suggest that not all strokes within a character have equal status during character identification, and a flexible stroke encoding system must underlie successful character identification during Chinese reading.

Multi-Font Farsi/Arabic Isolated Character Recognition Using Chain Codes

Nowadays, OCR systems have got several applications and are increasingly employed in daily life. Much research has been done regarding the identification of Latin, Japanese, and Chinese characters. However, very little investigation has been performed regarding Farsi/Arabic characters recognition. Probably the reason is difficulty and complexity of those characters identification compared to the others and limitation of IT activities in Farsi and Arabic speaking countries. In this paper, a technique has been employed to identify isolated Farsi/Arabic characters. A chain code based algorithm along with other significant peculiarities such as number and location of dots and auxiliary parts, and the number of holes existing in the isolated character has been used in this study to identify Farsi/Arabic characters. Experimental results show the relatively high accuracy of the method developed when it is tested on several standard Farsi fonts. Keywords—Farsi characters, OCR, feature extraction, chain code.

Use of phonological codes for Chinese characters: Evidence from processing of parafoveal preview when reading sentences

The role of phonological coding for character identification was examined with the benefit of processing parafoveal characters in eye fixations while reading Chinese sentences. In Experiment 1, the orthogonal manipulation of phonological and orthographic similarity can separate two types of phonological benefits for homophonic previews, according to whether these previews share the same phonetic radical with the targets or not. The significant phonological benefits indicate that phonological coding is activated early when the character is in the parafovea. Experiment 2 manipulated the character's consistency value and found that the phonological preview benefits are reliable only when the targets are high consistency characters. The results of two experiments suggest that phonological computation is rapid and early at both character and radical levels for Chinese character identification.

The role of phonological codes in integrating information across saccadic eye movements in Chinese character identification.

Prior research has generally assumed either that phonological codes do not contribute to Chinese character identification or that they do so only through a look-up process at the character level. In 3 experiments, a homophone seen parafoveally aided the identification of a target character that was fixated following an eye movement to the preview location. Moreover, high-frequency phonetically regular characters were named faster than high-frequency, phonetically irregular characters. Thus, both lexical and sublexical phonological codes of Chinese characters are involved early in the process of character identification. Orthographic information from the preview was also used in character identification, as orthographically similar previews facilitated target identification as well. The evidence for the extraction of semantic information from parafoveal previews was mixed, as synonym previews facilitated in Experiment 2 but not in Experiment 1.

The role of phonological codes in integrating information across saccadic eye movements in Chinese character identification.

Prior research has generally assumed either that phonological codes do not contribute to Chinese character identification or that they do so only through a look-up process at the character level. In 3 experiments, a homophone seen parafoveally aided the identification of a target character that was fixated following an eye movement to the preview location. Moreover, high-frequency phonetically regular characters were named faster than high-frequency, phonetically irregular characters. Thus, both lexical and sublexical phonological codes of Chinese characters are involved early in the process of character identification. Orthographic information from the preview was also used in character identification, as orthographically similar previews facilitated target identification as well. The evidence for the extraction of semantic information from parafoveal previews was mixed, as synonym previews facilitated in Experiment 2 but not in Experiment 1.