Kanji handriting recognition on Android

English version for iOS (iPhone, iPod Touch, iPad)

Asahi Kanji application for Android™
Single kanji Handwriting Recognition

This feature is obsolete and has been removed from the Asahi Kanji app in September 2025.
This was an experimental addition to the app, available between 2013 and 2025.
Handwriting recognition was based on the Tesseract OCR engine.

For a presentation of the whole application please see:
http://www.japanese-kanji.com/android/english/asahikanji-en-android.htm

The following presentation is kept as an archive.

How to use:

Trace a single kanji.
The thin box is only a guide to help you center the character and to suggest an amount of white space you can leave around it.
Do not worry if you write outside the box as its sides will not crop what you trace.
The image of the kanji will be analyzed by the Optical Recognition Engine (Tesseract) so stroke order has no importance here. You can even retouch a character to close a gap or make a stroke longer.

For a better recognition rate:

1) trace the kanji as big as you can.
2) contrarily to handwriting recognition based on strokes and their order, this approach does not give good result with free style calligraphy.
The character you trace must be as standard and scholarly as possible, like those used on public signs.
Do not trace the serif (the small upward stroke left by a brush at the end of a stroke.)
Do not leave gaps between strokes that are supposed to be joined.
The thickness of the strokes, relative to the size of the screen, seems to be the one that currently gives the best recognition rate.
We understand that tracing kanji with a high number of strokes with one finger is not easy. It is easier with a stylus.

After tracing the character, tap the "Analyze" button.

Limitations

Do not give up too quickly.
Just like the OCR feature you have to learn how to get the best of it.
Although counter intuitive, recognizing a single character can be a greater challenge for an OCR engine than recognizing a word made up of two kanji. Identifying a basic kanji made up of a few strokes is sometimes more difficult than recognizing one with a bigger number of strokes.
For example the Tesseract OCR engine cannot make the difference between 囗 and 口 , whatever the size of the drawn square on the screen.
The first one is a big square. It is the radical you find in the kanji 四 (four).
The second one, a smaller square, is the kanji "kuchi" (mouth, opening).
There are a few other similar cases, like 日 (sun), as Tesseract almost always returned the radical 曰 (the central horizontal stroke is a little shorter). For basic shapes Tesseract seems to favor radicals over kanji.
This is the reason why, until a better solution is found, the search results are limited to the Jōyō kanji.
For handwriting recognition all other characters and radicals (bushu) have been blacklisted.
Before that, 100% of the time, trying to display the card for 口 would display the message "Sorry, not a Jōyō kanji" or "Sorry, 曰 is not a Jōyō kanji", which is true. But we can imagine the comments on Google Play...

Your feedback, whether positive or negative, will be essential for the future development of this feature.

Credits:

The OCR module uses the Tesseract OCR engine, released under the Apache License, Version 2.0

Top of the page