Efficient mobile phone Chinese optical character recognition systems by use of heuristic fuzzy rules and bigram Markov language models

Cheok, Adrian David; Zhang, Jian; Chng, Eng Siong

doi:10.1016/j.asoc.2007.02.013

Cited by 9 publications

(3 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One of the most significant function of this system is that, the program uses TTS function to output ai as speech, which provides a more comfortable way for the blind to use the dictionary. In addition, several works in the past focused on OCR in android applications [35,36], real-time OCR [37], character readability on smart phones [38], character recognition models suitable for handheld devices [39], and App to recognize food items in Chinese menu [40]. Considering several related work, it is evident that, none of the previous research focused on developing a method to visually understand the text only by scanning.…”

Section: Related Workmentioning

confidence: 99%

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

Rai

2021

Complexity

View full text Add to dashboard Cite

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.

show abstract

Section: Related Workmentioning

confidence: 99%

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

Rai

2021

Complexity

View full text Add to dashboard Cite

show abstract

“…Unfortunately, such language models still have large memory complexity. Hence, they are inappropriate for use in embedded systems, e.g., mobile phones, which usually have less computational resources for modern applications (for example see in [12]) when compared to modern graphical processor units (GPU). This problem is especially challenging for the language modeling task, in which the RNNs are characterized by very large dimensionality of the last fully connected layer [13] because this layer produces |V| ≫ 1 posterior probabilities of all words from the vocabulary.…”

Section: Introductionmentioning

confidence: 99%

Compression of recurrent neural networks for efficient language modeling

Grachev

Ignatov

Savchenko

2019

Applied Soft Computing

View full text Add to dashboard Cite

Recurrent neural networks have proved to be an effective method for statistical language modeling. However, in practice their memory and run-time complexity are usually too large to be implemented in real-time offline mobile applications. In this paper we consider several compression techniques for recurrent neural networks including Long-Short Term Memory models. We make particular attention to the high-dimensional output problem caused by the very large vocabulary size. We focus on effective compression methods in the context of their exploitation on devices: pruning, quantization, and matrix decomposition approaches (low-rank factorization and tensor train decomposition, in particular). For each model we investigate the trade-off between its size, suitability for fast inference and perplexity. We propose a general pipeline for applying the most suitable methods to compress recurrent neural networks for language modeling. It has been shown in the experimental study with the Penn Treebank (PTB) dataset that the most efficient results in terms of speed and compression-perplexity balance are obtained by matrix decomposition techniques.

show abstract

“…Besides, the extensive applications of character recognition in recognizing the numbers on bank checks [2], car plate numbers [3], etc., have caused varieties of new systems, such as the OCR system to be developed [4]. In character recognition references, several different methods, such as the decision tree [5,6], fuzzy set theory [7][8][9][10], artificial neural networks [11,12], support vector machines [13,14], hidden Markov model [15][16][17][18] or any other significant hybrid of these methods [19,20] are exploited.…”

Section: Introductionmentioning

confidence: 99%