README.md

Handwriting touchpad

This project is about making use of touchpad for handwriting. It is heavily inspired by the Apple Trackpad on MacBook. As a poor undergraduate student who can't afford a MacBook, I decided to build a software that can turn my touchpad into a device that can be used for handwriting.

Why would you want to do handwriting on a computer? Can you just type the word?

I am a fan of practicing writing those crazy difficult Japanese Kanji. As a Japanese learner, you will not know how to type those words presented in Kanji. There is some words (e.g. people names) that even if you know the reading but when you type the reading (e.g. with Microsoft IME), it will never give you the kanji that you want.

Goal

• Retrieve absolute touchpad touches' positions
• Block mouse movement while using trackpad for handwriting. We use Windows Hooks to block mouse input (#5). Thanks to @alvinhochun.
• Visualize touch input position on GUI
• Integrate with TensorFlow (threading)

User guide

• The application can only be focused using Alt+Tab.
• Press F3 to enter touchpad writing mode. CAUTION Your mouse movement will be locked while you are in touchpad writing mode.
• Press ESC to exit touchpad writing mode.
• Press c to clear the canvas.
• The application currently can only be resized consistently with Windows Key + Arrow Keys as the application does not handle clicking properly.

Current progress

• We have been able to get the absolute touchpad input positions and visualize them on a UI window.
• We are having some problem with rendering the UI because the rendering workload is too heavy and the UI is re-rendered for every touch input events.

CI status

Terminology

• digitizer, touchpad, and trackpad are refering to the same things in this project.
• (a) touch and (a) contact are refering to the representation of the data generated when your finger(s) come(s) into contact with the hardware - the touchpad.
• HID - human interface device - refers to devices used by human to "communicate" with computers (even though they can be used by bot).

Progress

• We want to get the figure's position when it comes into contact with the touchpad (hereby refered as raw input data).
• Windows will send raw touchpad input data to GUI desktop application. (I am not sure how to get raw HID input from a console application.)
• Most of the APIs are for C/C++ so I decided to use C/C++ to create native Windows Desktop Application.
• We need to register our Desktop App for raw HID input so that Windows will send us the raw HID input data.
• Registering for raw HID input data is done in WndProc(HWND, UINT, WPARAM, LPARAM) function.
• WndProc is the event handler for Windows Desktop Application.
• We can register for raw HID input data when receiving WM_CREATE message (event).
• In Windows Desktop Application programing, the application receives messages from Windows and processes them in a callback function - WndProc in our case.
• In order to register for raw HID input data, we only need to call RegisterRawInputDevices(PCRAWINPUTDEVICE, UINT, UINT) once.
• We want Windows Precision Touchpad devices so our RAWINPUTDEVICE entry to be passed into RegisterRawInputDevices will look like this

RAWINPUTDEVICE rid;

rid.usUsagePage = HID_USAGE_PAGE_DIGITIZER;
rid.usUsage     = HID_USAGE_DIGITIZER_TOUCH_PAD;
rid.dwFlags     = RIDEV_INPUTSINK;
rid.hwndTarget  = hWnd;

• After registering for raw HID input data, Windows will send us raw input data for all Windows Precision Touchpad devices.

• If the device is connected after our register process, Windows will still send that device's raw HID input data to our Desktop Application so there is no need to calling RegisterRawInputDevices to get raw HID input data of devices that just has been connected.

• However, we may still need to handle multiple devices.

• The raw HID input data will be availabled with the WM_INPUT message (event).

• We can call GetRawInputData(HRAWINPUT, UINT, LPVOID, PUINT, UINT) to get either header or raw input data of the device.

• After getting raw input data, it will be in form of a byte pointer. In order to parse the raw input data, we need to do some processing with the preparsed data first.

• The preparsed data can be retrieved with the GetRawInputDeviceInfo(HANDLE, UINT, LPVOID, PUINT) call.

• After getting preparsed data, we have some helper functions from hid.lib to get information from the data.

• We have to include the hidusage.h and hidpi.h headers in mentioned order. hidpi.h requires some definitions from hidusage.h but it doesn't include the hidusage.h header. clang-format will sort headers and put hidpi.h on top of hidusage.h which is a bad thing. They also require Windows.h header and they also do not include it.

• We also need to add the hid.lib to additional dependencies in the Visual Studio project properties file. If you have multiple target platforms (e.g. x86 and x64) or multiple build configurations (e.g. Debug and Release), you will have to manually add the hid.lib to all of them.

• With some methods from hid.lib, we can get the following information:

  • the touchpad size (width and height)
  • maximum number (the number - as all the touch data is) of touches per raw input message
  • the link collection (ID) to get the tip switch value in each raw input message (to tell if the finger is currently on the touchpad surface or the finger is lifted from the touchpad surface)
  • the link collection for each touches (contacts) to get/locate the touch's data (position, tip switch, etc.) in the raw input message

• There is no convenient functions (e.g. GetDeviceWidth()) to retrieve those information. We have to call HidP_GetValuesCaps to get a array of device's capabilities. And we have to iterate through them to check their HID Usage and UsagePage to see if they contain the data (e.g. the width of the device) that we want.

• However, that array doesn't contain all the information that we need. There is another array called button capabilities from the HidP_GetButtonCaps API call.