multitouch/docs/report.tex

\documentclass[twoside,openright]{uva-bachelor-thesis}

\usepackage[english]{babel}
\usepackage[utf8]{inputenc}
\usepackage{hyperref,graphicx,float,tikz,subfigure}

% Link colors
\hypersetup{colorlinks=true,linkcolor=black,urlcolor=blue,citecolor=DarkGreen}

% Title Page
\title{A generic architecture for the detection of multi-touch gestures}
\author{Taddeüs Kroes}
\supervisors{Dr. Robert G. Belleman (UvA)}
\signedby{Dr. Robert G. Belleman (UvA)}

\begin{document}

% Title page
\maketitle
\begin{abstract}
    % TODO
\end{abstract}

% Set paragraph indentation
\parindent 0pt
\parskip 1.5ex plus 0.5ex minus 0.2ex

% Table of contant on separate page
\tableofcontents

\chapter{Introduction}

% Ruwe probleemstelling
Multi-touch interaction is becoming increasingly common, mostly due to the wide
use of touch screens in phones and tablets. When programming applications using
this method of interaction, the programmer needs an abstraction of the raw data
provided by the touch driver of the device. This abstraction exists in several
multi-touch application frameworks like Nokia's
Qt\footnote{\url{http://qt.nokia.com/}}. However, applications that do not use
these frameworks have no access to their multi-touch events.

% Aanleiding
This problem was observed during an attempt to create a multi-touch
``interactor'' class for the Visualization Toolkit \cite[VTK]{VTK}. Because VTK
provides the application framework here, it is undesirable to use an entire
framework like Qt simultaneously only for its multi-touch support.

% Ruw doel
The goal of this project is to define a generic multi-touch event triggering
architecture. To test the definition, a reference implementation is written in
Python.

    \section{Definition of the problem}

    % Hoofdvraag
    The goal of this thesis is to a create generic architecture for a
    multi-touch event triggering mechanism for use in multi-touch applications.

    % Deelvragen
    To design such an architecture properly, the following questions are relevant:
    \begin{itemize}
        \item What is the input of the architecture? Different touch drivers
            have different API's. To be able to support different drivers
            (which is highly desirable), there should be a translation from the
            driver API to a fixed input format.
        \item How can extendability be accomplished? The set of supported
            events should not be limited to a single implementation, but an
            application should be able to define its own custom events.
        \item How can the architecture be used by different programming
            languages? A generic architecture should not be limited to be used
            in only one language.
        \item Can events be used by multiple processes at the same time? For
            example, a network implementation could run as a service instead of
            within a single application, triggering events in any application
            that needs them.
    \end{itemize}

    % Afbakening
    The scope of this thesis includes the design of a generic multi-touch
    triggering architecture, a reference implementation of this design, and its
    integration into a test case application. To be successful, the design
    should allow for extensions to be added to any implementation.

    The reference implementation is a Proof of Concept that translates TUIO
    events to some simple touch gestures that are used by some test
    applications.
    %Being a Proof of Concept, the reference implementation itself does not
    %necessarily need to meet all the requirements of the design.

    \section{Structure of this document}

    % TODO: pas als thesis af is

\chapter{Related work}

    \section{Gesture and Activity Recognition Toolkit}

    The Gesture and Activity Recognition Toolkit (GART) \cite{GART} is a
    toolkit for the development of gesture-based applications. The toolkit
    states that the best way to classify gestures is to use machine learning.
    The programmer trains a program to recognize using the machine learning
    library from the toolkit. The toolkit contains a callback-mechanism that
    the programmer uses to execute custom code when a gesture is recognized.

    Though multi-touch input is not directly supported by the toolkit, the
    level of abstraction does allow for it to be implemented in the form of a
    ``touch'' sensor.

    The reason to use machine learning is the statement that gesture detection
    ``is likely to become increasingly complex and unmanageable'' when using a
    set of predefined rules to detect whether some sensor input can be seen as
    a specific gesture. This statement is not necessarily true. If the
    programmer is given a way to separate the detection of different types of
    gestures and flexibility in rule definitions, over-complexity can be
    avoided.

    % oplossing: trackers. bijv. TapTracker, TransformationTracker gescheiden

    \section{Gesture recognition software for Windows 7}

    % TODO
    The online article \cite{win7touch} presents a Windows 7 application,
    written in Microsofts .NET. The application shows detected gestures in a
    canvas. Gesture trackers keep track of stylus locations to detect specific
    gestures. The event types required to track a touch stylus are ``stylus
    down'', ``stylus move'' and ``stylus up'' events. A
    \texttt{GestureTrackerManager} object dispatches these events to gesture
    trackers. The application supports a limited number of pre-defined
    gestures.

    An important observation in this application is that different gestures are
    detected by different gesture trackers, thus separating gesture detection
    code into maintainable parts.

    \section{Processing implementation of simple gestures in Android}

    An implementation of a detection architecture for some simple multi-touch
    gestures (tap, double tap, rotation, pinch and drag) using
    Processing\footnote{Processing is a Java-based development environment with
    an export possibility for Android. See also \url{http://processing.org/.}}
    can be found found in a forum on the Processing website
    \cite{processingMT}. The implementation is fairly simple, but it yields
    some very appealing results. The detection logic of all gestures is
    combined in a single class. This does not allow for extendability, because
    the complexity of this class would increase to an undesirable level (as
    predicted by the GART article \cite{GART}). However, the detection logic
    itself is partially re-used in the reference implementation of the
    generic gesture detection architecture.

    \section{Analysis of related work}

    The simple Processing implementation of multi-touch events provides most of
    the functionality that can be found in existing multi-touch applications.
    In fact, many applications for mobile phones and tablets only use tap and
    scroll events. For this category of applications, using machine learning
    seems excessive. Though the representation of a gesture using a feature
    vector in a machine learning algorithm is a generic and formal way to
    define a gesture, a programmer-friendly architecture should also support
    simple, ``hard-coded'' detection code. A way to separate different pieces
    of gesture detection code, thus keeping a code library manageable and
    extendable, is to user different gesture trackers.

\chapter{Requirements}

% testimplementatie met taps, rotatie en pinch. Hieruit bleek:
% - dat er verschillende manieren zijn om bijv. "rotatie" te
%   detecteren, (en dat daartussen onderscheid moet kunnen worden
%   gemaakt)
% - dat detectie van verschillende soorten gestures moet kunnen
%   worden gescheiden, anders wordt het een chaos.
% - Er zijn een aantal keuzes gemaakt bij het ontwerpen van de gestures,
%   bijv dat rotatie ALLE vingers gebruikt voor het centroid. Het is
%   wellicht in een ander programma nodig om maar 1 hand te gebruiken, en
%   dus punten dicht bij elkaar te kiezen (oplossing: windows).

    \section{Introduction}

    % TODO
    TODO: doel v/h experiment

    To test multi-touch interaction properly, a multi-touch device is required.
    The University of Amsterdam (UvA) has provided access to a multi-touch
    table from PQlabs. The table uses the TUIO protocol \cite{TUIO} to
    communicate touch events. See appendix \ref{app:tuio} for details regarding
    the TUIO protocol.

    \section{Experimenting with TUIO and event bindings}
    \label{sec:experimental-draw}

    When designing a software library, its API should be understandable and
    easy to use for programmers. To find out the basic requirements of the API
    to be usable, an experimental program has been written based on the
    Processing code from \cite{processingMT}. The program receives TUIO events
    and translates them to point \emph{down}, \emph{move} and \emph{up} events.
    These events are then interpreted to be (double or single) \emph{tap},
    \emph{rotation} or \emph{pinch} gestures. A simple drawing program then
    draws the current state to the screen using the PyGame library. The output
    of the program can be seen in figure \ref{fig:draw}.

    \begin{figure}[H]
        \center
        \label{fig:draw}
        \includegraphics[scale=0.4]{data/experimental_draw.png}
        \caption{Output of the experimental drawing program. It draws the touch
                 points and their centroid on the screen (the centroid is used
                 as center point for rotation and pinch detection). It also
                 draws a green rectangle which responds to rotation and pinch
                 events.}
    \end{figure}

    One of the first observations is the fact that TUIO's \texttt{SET} messages
    use the TUIO coordinate system, as described in appendix \ref{app:tuio}.
    The test program multiplies these with its own dimensions, thus showing the
    entire screen in its window. Also, the implementation only works using the
    TUIO protocol. Other drivers are not supported.

    Though using relatively simple math, the rotation and pinch events work
    surprisingly well. Both rotation and pinch use the centroid of all touch
    points. A \emph{rotation} gesture uses the difference in angle relative to
    the centroid of all touch points, and \emph{pinch} uses the difference in
    distance. Both values are normalized using division by the number of touch
    points. A pinch event contains a scale factor, and therefore uses a
    division of the current by the previous average distance to the centroid.

    There is a flaw in this implementation. Since the centroid is calculated
    using all current touch points, there cannot be two or more rotation or
    pinch gestures simultaneously. On a large multi-touch table, it is
    desirable to support interaction with multiple hands, or multiple persons,
    at the same time. This kind of application-specific requirements should be
    defined in the application itself, whereas the experimental implementation
    defines detection algorithms based on its test program.

    Also, the different detection algorithms are all implemented in the same
    file, making it complex to read or debug, and difficult to extend.

    \section{Summary of observations}
    \label{sec:observations}

    \begin{itemize}
        \item The TUIO protocol uses a distinctive coordinate system and set of
            messages.
        \item Touch events occur outside of the application window.
        \item Gestures that use multiple touch points are using all touch
            points (not a subset of them).
        \item Code complexity increases when detection algorithms are added.
        \item A multi-touch application can have very specific requirements for
            gestures.
    \end{itemize}

    \section{Requirements}
    \label{sec:requirements}

    From the observations in section \ref{sec:observations}, a number of
    requirements can be specified for the design of the event mechanism:

    \begin{itemize}
        % vertalen driver-specifieke events naar algemeen formaat
        \item To be able to support multiple input drivers, there must be a
            translation from driver-specific messages to some common format
            that can be used in gesture detection algorithms.
        % events toewijzen aan GUI window (windows)
        \item An application GUI window should be able to receive only events
            occurring within that window, and not outside of it.
        % scheiden groepen touchpoints voor verschillende gestures (windows)
        \item To support multiple objects that are performing different
            gestures at the same time, the architecture must be able to perform
            gesture detection on a subset of the active touch points.
        % scheiden van detectiecode voor verschillende gesture types
        \item To avoid an increase in code complexity when adding new detection
            algorithms, detection code of different gesture types must be
            separated.
        % extendability
        \item The architecture should allow for extension with new detection
            algorithms to be added to an implementation. This enables a
            programmer to define custom gestures for an application.
    \end{itemize}

% -------
% Results
% -------

\chapter{Design}

    \section{Components}

        Based on the requirements from section \ref{sec:requirements}, a design
        for the architecture has been created. The design consists of a number
        of components, each having a specific set of tasks.

        \subsection{Event server}

        % vertaling driver naar point down, move, up
        % vertaling naar schermpixelcoordinaten
        % TUIO in reference implementation

        The \emph{event server} is an abstraction for driver-specific server
        implementations, such as a TUIO server. It receives driver-specific
        messages and tanslates these to a common set of events and a common
        coordinate system.

        A minimal example of a common set of events is $\{point\_down,
        point\_move, point\_up\}$. This is the set used by the reference
        implementation. Respectively, these events represent an object being
        placed on the screen, moving along the surface of the screen, and being
        released from the screen.

        A more extended set could also contain the same three events for an
        object touching the screen. However, a object can also have a
        rotational property, like the ``fiducials'' type in the TUIO protocol.
        This results in $\{point\_down, point\_move, point\_up, object\_down,
        object\_move, object\_up,\\object\_rotate\}$.
        % TODO: is dit handig? point_down/object_down op 1 of andere manier samenvoegen?

        An important note here, is that similar events triggered by different
        event servers must have the same event type and parameters. In other
        words, the output of the event servers should be determined by the
        gesture servers (not the contrary).

        The output of an event server implementation should also use a common
        coordinate system, that is the coordinate system used by the gesture
        server. For example, the reference implementation uses screen
        coordinates in pixels, where (0, 0) is the upper left corner and
        (\emph{screen width}, \emph{screen height}) the lower right corner of
        the screen.

        The abstract class definition of the event server should provide some
        functionality to detect which driver-specific event server
        implementation should be used.

        \subsection{Gesture trackers}

        Like \cite[the .NET implementation]{win7touch}, the architecture uses a
        \emph{gesture tracker} to detect if a sequence of events forms a
        particular gesture. A gesture tracker detects and triggers events for a
        limited set of gesture types, given a set of touch points. If one group
        of touch points is assigned to one tracker and another group to another
        tracker, multiple gestures can be detected at the same time. For the
        assignment of different groups of touch points to different gesture
        trackers, the architecture uses so-called \emph{windows}. These are
        described in the next section.

        % event binding/triggering
        A gesture tracker triggers a gesture event by executing a callback.
        Callbacks are ``bound'' to a tracker by the application. Because
        multiple gesture types can have very similar detection algorithm, a
        tracker can detect multiple different types of gestures. For instance,
        the rotation and pinch gestures from the experimental program in
        section \ref{sec:experimental-draw} both use the centroid of all touch
        points.

        If no callback is bound for a particular gesture type, no detection of
        that type is needed. A tracker implementation can use this knowledge
        for code optimization.

        % scheiding algoritmiek
        A tracker implementation defines the gesture types it can trigger, and
        the detection algorithms to trigger them. Consequently, detection
        algorithms can be separated in different trackers. Different
        trackers can be saved in different files, reducing the complexity of
        the code in a single file. \\
        % extendability
        Because a tracker defines its own set of gesture types, the application
        developer can define application-specific trackers (by extending a base
        \texttt{GestureTracker} class, for example). In fact, any built-in
        gesture trackers of an implementation are also created this way. This
        allows for a plugin-like way of programming, which is very desirable if
        someone would want to build a library of gesture trackers. Such a
        library can easy be extended by others.

        \subsection{Windows}

        A \emph{window} represents a subset of the entire screen surface. The
        goal of a window is to restrict the detection of certain gestures to
        certain areas. A window contains a list of touch points, and a list of
        trackers. A gesture server (defined in the next section) assigns touch
        points to a window, but the window itself defines functionality to
        check whether a touch point is inside the window. This way, new windows
        can be defined to fit over any 2D object used by the application.

        The first and most obvious use of a window is to restrict touch events
        to a single application window. However, the use of windows can be used
        in a lot more powerful way.

        For example, an application contains an image with a transparent
        background that can be dragged around. The user can only drag the image
        by touching its foreground. To accomplish this, the application
        programmer can define a window type that uses a bitmap to determine
        whether a touch point is on the visible image surface. The tracker
        which detects drag gestures is then bound to this window, limiting the
        occurence of drag events to the image surface.

        % toewijzen even aan deel v/h scherm:
        % TUIO coördinaten zijn over het hele scherm en van 0.0 tot 1.0, dus
        % moeten worden vertaald naar pixelcoördinaten binnen een ``window''
        % TODO

        \subsection{Gesture server}

        % luistert naar point down, move, up
        The \emph{gesture server} delegates events from the event server to the
        set of windows that contain the touch points related to the events.

        % toewijzing point (down) aan window(s)
        The gesture server contains a list of windows. When the event server
        triggers an event, the gesture server ``asks'' each window whether it
        contains the related touch point. If so, the window updates its gesture
        trackers, which can then trigger gestures.

    \section{Diagrams}

    \input{data/diagrams}
    \simplediagram
    \completediagrams

    \section{Example usage}

    This section describes an example that illustrates the communication
    between different components. The example application listens to tap events
    in a GUI window.

    \begin{verbatim}
    # Create a gesture server that will be started later
    server = new GestureServer object

    # Add a new window to the server, representing the GUI
    window = new Window object
    set window position and size to that of GUIO window
    add window to server

    # Define a handler that must be triggered when a tap gesture is detected
    begin function handler(gesture)
        # Do something
    end function

    # Create a tracker that detects tap gestures
    tracker = new TapTracker object  # Where TapTracker is an implementation of
                                     # abstract Tracker
    add tracker tot window
    bind handler to tracker.tap

    # If the GUI toolkit allows it, bind window movement and resize handlers
    # that alter the position size and sieze of the window object

    # Start the gesture server (which in turn starts a driver-specific event
    # server)
    start server
    \end{verbatim}

    \section{\emph{hier moet een conslusie komen die de componenten aansluit op de requirements(?)}}

    % TODO
    %

    %\section{Network protocol}

    % TODO
    % ZeroMQ gebruiken voor communicatie tussen meerdere processen (in
    % verschillende talen)

\chapter{Reference implementation}

% TODO
% alleen window.contains op point down, niet move/up
% geen netwerk protocol
% een paar simpele windows en trackers

%\chapter{Conclusions}

% TODO
% Windows zijn een manier om globale events toe te wijzen aan vensters
% Trackers zijn een effectieve manier om gebaren te detecteren
% Trackers zijn uitbreidbaar door object-orientatie

\chapter{Suggestions for future work}

% TODO

% geruik formele definitie van gestures in gesture trackers, bijv. state machine

% Network protocol (ZeroMQ) voor meerdere talen en simultane processen
% Hierij ook: extra laag die gesture windows aanmaakt die corresponderen met window manager

% Window in boomstructuur voor efficientie

\bibliographystyle{plain}
\bibliography{report}{}

\appendix

\chapter{The TUIO protocol}
\label{app:tuio}

The TUIO protocol \cite{TUIO} defines a way to geometrically describe tangible
objects, such as fingers or objects on a multi-touch table. Object information
is sent to the TUIO UDP port (3333 by default).

For efficiency reasons, the TUIO protocol is encoded using the Open Sound
Control \cite[OSC]{OSC} format. An OSC server/client implementation is
available for Python: pyOSC \cite{pyOSC}.

A Python implementation of the TUIO protocol also exists: pyTUIO \cite{pyTUIO}.
However, the execution of an example script yields an error regarding Python's
built-in \texttt{socket} library. Therefore, the reference implementation uses
the pyOSC package to receive TUIO messages.

The two most important message types of the protocol are ALIVE and SET
messages. An ALIVE message contains the list of session id's that are currently
``active'', which in the case of multi-touch a table means that they are
touching the screen. A SET message provides geometric information of a session
id, such as position, velocity and acceleration.

Each session id represents an object. The only type of objects on the
multi-touch table are what the TUIO protocol calls ``2DCur'', which is a (x, y)
position on the screen.

ALIVE messages can be used to determine when an object touches and releases the
screen. For example, if a session id was in the previous message but not in the
current, The object it represents has been lifted from the screen.

SET provide information about movement. In the case of simple (x, y) positions,
only the movement vector of the position itself can be calculated. For more
complex objects such as fiducials, arguments like rotational position is also
included.

ALIVE and SET messages can be combined to create ``point down'', ``point move''
and ``point up'' events (as used by the \cite[.NET application]{win7touch}).

TUIO coordinates range from $0.0$ to $1.0$, with $(0.0, 0.0)$ being the left
top corner of the screen and $(1.0, 1.0)$ the right bottom corner. To focus
events within a window, a translation to window coordinates is required in the
client application, as stated by the online specification
\cite{TUIO_specification}:
\begin{quote}
    In order to compute the X and Y coordinates for the 2D profiles a TUIO
    tracker implementation needs to divide these values by the actual sensor
    dimension, while a TUIO client implementation consequently can scale these
    values back to the actual screen dimension.
\end{quote}

\end{document}