I have moved to new website, http://hsyeo.com

I have moved to new website, http://hsyeo.com , see you there!

Hand and Fingers tracking + Simple static gesture recognition, using webcam only

 Hand and Fingers tracking + Simple static gesture recognition, using webcam only

Progress update (End of December)

(Updating...)

• Received LeapMotion Developer Kit (22 Jan)
• Received Intel Perceptual Camera (11 Jan)
• Revised MTAP journal major revision (19 Jan)
• Submitted CUTE III journal camera ready version (~21 Dec)
• Attended CUTE conference in HK (22 Dec)
• Submitted a paper to Mobisys 2013 conference (~ 10 Dec)

About Hand tracking
-updated algorithm (K-curvature + convexity)
-added thumb finger, added left right hand differentiation
-added static gestures recognition (Open, Close, Palm, OK, O, Gun, Pinch, Claw)
-added TUIO support
-added finger direction
-added min enclosing circle for check palm openness
TODO:
-still thinking

About MultiCloud
-added Google Drive and Skydrive, now support 4 clouds including Dropbox and Box
-added split and join
-added erasure coding
-added database
-added encryption and compression, in pipeline
-added pipeline parallel processing
TODO:
-add range support with REST API, download different parts from mirrors
-add SugarSync, Amazon S3

WebDAV VSS
-doing multiple accounts without database (pref only)
-improving caching method, maybe need database
TODO:
-fix bugs (50%)
-shortcut to download folder
-notification bar
-list sorting, search functions
--stream any part, using range and dummy
--multi chunk dl from server, using range (50%)
--multi write into same file using java nio filechannel (80%, but slower)
--compression, terus unzip and open, or mount, using zip input stream (50%)
--database for store last modified date becoz gnex cannot modify lmdate
--database for storing whole list, and other useful stuffs (multi accs, ...)
--resumable download
---thumbnails
---SSL
---swipe left to open, right to download in listview
---AES encryption
---show status in list (cached, not cached, % health cached)
---learn from nexes file manager & others open source
---multi select
---change runnable to asynctask
----rsync?
----improve sardine code, try sardine new checkout in new project

Remote Downloader
-finalize & set for public download
-fix bugs
-AES->SSL
-C# vs java packet problem (use basic Object parsing? or XML or JSON?)

Multi-Clouds Storage
-add gdrive(still waiting example), sugar
-
TODO:
-split and join files
-download different part from different location (dropbox & box api no support range, need to use REST api zzz)

Progress update 07 october 2012

WebDAV VSS
-doing multiple accounts without database (pref only)
-improving caching method, maybe need database
TODO:
-fix bugs (50%)
-shortcut to download folder
-notification bar
-list sorting, search functions
--stream any part, using range and dummy
--multi chunk dl from server, using range (50%)
--multi write into same file using java nio filechannel (80%, but slower)
--compression, terus unzip and open, or mount, using zip input stream (50%)
--database for store last modified date becoz gnex cannot modify lmdate
--database for storing whole list, and other useful stuffs (multi accs, ...)
--resumable download
---thumbnails
---SSL
---swipe left to open, right to download in listview
---AES encryption
---show status in list (cached, not cached, % health cached)
---learn from nexes file manager & others open source
---multi select
---change runnable to asynctask
----rsync?
----improve sardine code, try sardine new checkout in new project

Remote Downloader
-finalize & set for public download
-fix bugs
-AES->SSL
-C# vs java packet problem (use basic Object parsing? or XML or JSON?)

Multi-Clouds Storage
-add gdrive(still waiting example), sugar
-
TODO:
-split and join files
-download different part from different location (dropbox & box api no support range, need to use REST api zzz)

Progress update 10 September 2012

WebDAV VSS
-doing multiple accounts without database (pref only)
-improving caching method, maybe need database
TODO:
-fix bugs
-shortcut to download folder
-notification bar
-list sorting, search functions
--stream any part, using range and dummy
--multi chunk dl from server, using range
--compression, terus unzip and open, or mount, using zip input stream
---rsync?
---improve sardine code

Multi-Clouds Storage
-add gdrive, sugar
-
TODO:
-split and join files
-download different part from different location

Progress update 12 August

15/7-15/8
  - WebDAV app upgrade (database, remote dler packet & SSL) (upgrading)
  - MultiCloud project (start?) (started)
  - Fyp paper for local journal (submitted to MTAP journal)

  - Hand tracking for international journal (no time to write, feel like not much contribution)

2/8-12/8 
Multi Cloud started
- able to connect 2 clouds, perform authentication
- show user information
- list files in 2 clouds (in single listview)
- simple download, delete
 TODO:
+ copy, move, upload
+ cache, compare cache
+ media streaming using stream proxy
+ split and join files

WebDAV app upgrading
- added ads
- try to publish in market
- fix some bugs
- improve local file copy, move functions
TODO:
+ fix more bugs
+ improve caching method
+ fix remote downloader (client and server side)
+ fix sardine

Next??
Prof lee want me to explain the hand tracking method
- I will suggest more advance multi touch table

Blind type Kinect Virtual Keyboard?
- No time to work on this...

Design simple online virtual file system
- Using FUSE? FuseAndroid?
- What is cross compiling the source?
- I need to study some linux..


Pledged for Oculus Rift in Kickstarter
Funded Ouya in Kickstarter


16/8 Back to Msia
29/8 Back to Busan

Progress update 22/july

Low output recently, keep studying/researching about good topic but to no avail
some nice storage research websites here, but mostly suitable for enterprise only
http://www.techtarget.com/
http://www.wired.com/cloudline/
http://www.storage-switzerland.com/Articles/Archive.html
http://www.enterprisestorageforum.com/
http://www.infostor.com/index.html
http://www.datacenterknowledge.com/
http://www.gartner.com/technology/home.jsp



15/7-15/8
  - WebDAV app upgrade (database, remote dler packet & SSL)
  - MultiCloud project (start?)
  - Fyp paper for local journal
  - Hand tracking for international journal

16/8 Back to Msia
29/8 Back to Busan

NEED NEW IDEA FOR SCI/SCIE LEVEL PLEASE!!

New way of typing on mutlti-touch tablet (just idea)

EDITED 2 (24/7/2012) Heh heh heh hee hee hu hu ha ha (joker's laugh)
So last week I thought of an idea about the BlindType, but then I immediately realized that it has been done about 2 years ago.
But now I am thinking of implementing the blind type technique into my finger tracking kinect application. So it would make typing in the air possible, without good accuracy.
Please dont KS me anymore, let me SCIE this time!
EDITED: After I blogged about this idea, then I start google about it if anyone has done it before.
And... yup, there is existing application doing the same thing, and it seems very good.
So... I am outdated... for about 2 years...
Please refer to the link http://www.androidpolice.com/2010/08/08/blindtype-keyboard-to-come-to-android-type-anywhere-on-your-phones-screen-even-with-no-keyboard-present/
But blindtype never gets to the public, it was bought by google, and google has been implement the technology in android 4.0
However there is still 1 thing google didnt implemented, the main power of blindtype, which is type anywhere on the screen.

Typing on smartphone is hard, most of people rely on predictive keyboard (T9, swype, swiftkey)
On smartphone, I personally favor swype although bunch of people say swiftkey tends to be very intelligent over time, and it can predict most of your next words.

But well, on tablet, swype isn't a very good input method. As you need to swype through a very long distance, it is very tired..

While the screen on tablet is bigger, people tends to go back to the original = typing on the on-screen keyboard. Because the area is big enough, and each key is individual enough, it can simulate the typing on keyboard action.

However, it is not true at all. The experience is still very much different from typing on a real keyboard, it is not even close! I am not saying about the tactile feedback here, nor the hand feeling.

User still need to look at the on screen keyboard on the tablet when they are typing, it makes the typing experience sucks, and it slows down the WPM. I am not going to explain deeply on the experience here.

When we type on a real keyboard, we don't need to see the keyboard at all. (for average users and above, not for noobs)
What if we can simulate the same experience on tablet?

So my idea is:
We can type randomly on the tablet screen (I mean anywhere), the same way we did on normal keyboard.
Then a software will capture all the points, and predict the word we type.
Don't understand? Let see some picture.
For example, when u type "l o v e r", as shown in next picture
your finger land on the location by sequence(highlighted in red color)
we assume that the location where your finger lands, will be almost the same for a particular alphabet, everytime






 lets remove the keyboard background and see what we have there


So we got some unknown locations, in sequence
and some relative distance for each sequence

And we also have relative distance for next n-th sequence for key 1 correspond to key 3, key 4 or key 5

With so much information gathered, we can predict the word we just typed

When u start pressing first key, the software recognize the location, then when u type the second key, the software recognize the location, and calculate the relative distance to the first point. And so on...
finally it predict a word based on the collected points.

Remembered when I say you can type on anywhere on the tablet screen?
let see some example if it is possible



versus


Even if you type anywhere, the relative distance, and sequence will be ALMOST the same for each particular word. So, it doesn't matter. The intelligent software will predict your word by calculating the location, relative distance, and etc.
So I can say it is very possible and feasible.


Since swype and swiftkey is so clever, it must be possible.
swype can predict a word based on the keys u swipe through, while swifkey can predict your word even you only type the 1st alphabet, and even predict ur next word.
Combining swype, swiftkey with my idea, then it makes perfect typing experience on tablet!

Above was just my little tiny idea.
I cannot make it possible due to lack of skills especially in text prediction.