The goal of this research is to create a modeling environment for non-expert users to allow them to easily create applications for mobile health monitoring. In this paper, we focus on the design and representation of the modeling constructs that represent information taken from wearable sensors. We want to find a way to represent sensor information such that it is not confusing and is easily recognizable by the non-expert user.
Model-driven development is a widely accepted approach
to simplifying the development of complex systems. This
involves the development of applications by creating a
highlevel model of the application and automatically transforming
this high level model to platform specific code [
End-user creation of mobile health monitoring applications can be beneficial to individuals, medical institutions and research groups that focus on collecting biological data from individuals. These types of people usually have little or no experience at all in programming such systems, and therefore need either the help of some programmer or have to learn programming themselves which is not an easy task. The following are some example scenarios that illustrate the benefits of end-user creation of mobile health monitoring applications. Suppose a person wants to have a health monitoring application that would keep track of how much he spends on physical activities with the use of a wearable device that records movements and transmits movement information to a mobile phone. However, he wants some customized view of the software and additional functionalities that are not present in the vendor-provided software of the wearable sensor. He could just use a modeling tool that would allow him to easily create this specialized application based on his wishes, and in turn, this tool will create the application for him without him having to worry about low-level programming. Another application is in clinics that provide patients with health monitoring devices that keep Heinrich Hussmann Department of Computer Science
The advancement of technologies has allowed the creation of
small wearable devices that are relatively non-intrusive, but
which can measure biological data in real-time and allows the
transmission of the collected data through wireless means (e.g.
transmission through Bluetooth). Some example commercial
systems and researches which applies such technologies are the
Nike+iPod Sports Kit [
level data sent by the device by providing interfaces that
represent the information in a more human-understandable form.
For instance, the BodyMovement class [
For our research, we have taken the Alive Technologies devices as the example target wearable device and the Java Mobile Edition (J2ME) framework combined with the APIs provided by MobHealth as the target code. In the next section we will first give an overview of our conceptual modeling framework and briefly describe the modeling tool that we are currently developing. We will then discuss in section 4 how we propose to model sensor data.
The Mobile Applications (Mobia) Modeling Framework is a framework, which we are currently designing and developing in order to simplify the development of applications for mobile platforms. We introduce one module of Mobia which concentrates on the development of applications for health monitoring. These types of applications involve wearable sensors that collect information from a person, and a mobile phone which processes the data taken from the sensor. Applications that involve separate devices for collecting information are typically difficult to develop, that is why we propose to apply the model-driven approach to the creation of such applications. However, since the target users of framework are non-expert users, we want to create a modeling environment that is easily learnable and features modeling constructs that are intuitive enough. We are currently developing a prototype called Mobia Modeler Suite, which is a tool that allows the simple graphical modeling of mobile applications. Since we target non-experts to be the main users of this tool, we must find ways in order to present information in a clear manner. The need for developing several prototypes of the interface for the modeling environment is important in order for us to evaluate the different possible ways in presenting the interface to the user.
For this paper, we focus on discussing the design issues involved in visually representing sensor data in the Mobile Health Monitoring Module of the Mobia Modeler Suite. More about this will be discussed in the next section.
The importance of trying to represent sensor data in a modeling environment is that, we want users to know what kind of information they can expect from a certain device and help them to be able to immediately visualize what they expect to appear on their applications. Too many technical details about a certain sensor should be abstracted from the user of the modeling tool. We must provide a way to give the user what they initially expect to get from a certain device, but at the same time provide additional information that the user has not initially thought about, but could probably use in their applications.
We have introduced the medget modeling construct to represent a certain device in the model. This represents a wearable sensor that can measure specific aspects of health. For research purposes, we focus on three types of medgets: the thermometer, the ECG meter and the actimeter, which measures temperature, ECG and movement respectively. We also assume here that each medget only functions as one device that measures only one specific type of bodily function, although in reality, a medget may measure multiple bodily functions (e.g. Alive Heart Monitor which not only measures ECG, but movement as well through accelerometer signals).
The following points are the things we want to identify in this research: •
• •
How to show all the available medgets in the modeling environment in a manner that it is not confusing to the user.
medget to the screen.
In order to find out the answers to the points mentioned, we made several designs for the medget constructs and a design evaluation in a form of a survey was made. There were 14 participants to the survey, 8 of which have backgrounds in Computer Science while the others are from Engineering, Mathematics and Physics. The evaluation made was more of a subjective and qualitative evaluation, relying on the participant’s previous knowledge and expertise. Details on the approaches we made will be discussed in the three subsections.
One of the concerns in the design of the Medget is that, what type of information should be available to the user in the modeling environment. We have identified three types of information that could be presented to the user: the name of the medget, a symbol representing it and the data it can provide. For ease of explanation purposes, let’s take the Thermometer medget as a concrete example. Table 1 shows the three types of information that can be provided by a medget, and a certain instance of it, which is the thermometer. We want to know here, which information matters most to the user. Figure 2 shows all the possible combinations of information that we have presented to the user, and the percentage of which the users have chosen. Take note here that few of the users indicated multiple choices. A conclusion drawn from the said result is that, the user is not too concerned with the name of the medget as long as the symbols representing it is clear enough for the user. Textual labels were suggested to be shown during a mouseover event for instance. Another thing is the importance of showing the data that a medget provides e.g. Celcius for the Thermometer.
Another thing we wanted to find out is with regards to how we can show all the available medgets and the data they provide in the modeling environment. For this example, let us assume that there are three types of medgets available: medget A, medget B and medget C. We will not name the medgets here since what is important is how they are positioned and displayed, not particularly the information they contain. Now, each medget can have 1 or more types of data that it can provide which are shown as cross symbols. For the Individual display shown in Figure 3, each medget and all possible data it provides is shown as one component in the medget palette. For instance, if there are three available data type for each medget, then three components for that medget are present in the palette. The Grouped display shown in Figure 4 shows all the possible data grouped according to the medget it is taken from. The medget palette is divided into three sections for the three types of medget, and each section contains all the possible data. The Minimalist display in Figure 5 shows a very minimal representation of medgets in the palette (box on the right). Later on, when the device is attached to a mobile screen (box on the left), double clicking on the medget icon on top of the screen will show a possible data representation for each double click. Based on the user survey, 46% chose the Grouped display, while 31% chose the minimalist and the rest chose the individual display. The focus is first on the medget that sends out the information and then later on, to the type of data it provides. It is also easier to look for what the users need when things are grouped. One of the design questions that we wanted to answer when trying to visualize information for applications in mobile health monitoring is that, how can we try to represent the flow of information from one medget to the screen? For this one, we have provided 3 different types of techniques. The first one is the Choose, Drag and Drop Approach which is the common approach in programming or modeling environments. Given a set of components in the palette, the user chooses the component and then drags it to the target location. The second one is the drag, connect and click approach, in which the user drags an arrow from the medget to the target screen. This arrow will signify information would flow from the medget to the screen. After a connection between a screen and a medget is done, the user would then click on the desired data that would appear on the mobile screen.
Finally, the Drag, Drop and Click is an approach in which the user drags a certain medget to the target screen and the medget is attached to the top of the screen to symbolize connection. The default data provided by a medget will then be shown on the screen. Clicking on the attached medget will show a possible data representation for each double click. For instance, if a medget has two available data representations, the visuals will toggle between the two each time a double click is done. Based on the survey results, majority of the participants preferred the Choose, Drag and Drop approach (46%). A big factor that may have resulted from this is because it is the concept that the participants were most familiar with. Second choice (38%) was the drag, drop and click approach. Although the least popular which was the drag, connect and click approach actually symbolizes some sort of connection with the medget in the presence of flow arrows in comparison to the real world, the steps involved seems too deviate from existing approaches. In conclusion, although the other methods can be learned easily, the most commonly encountered approach was still the most preferred among the participants.
We have presented in this paper our ongoing work on creating a user-friendly modeling environment for modeling applications for mobile health monitoring. The main core of this paper was to discuss how we can model sensor data taken from wearable sensors. We presented some design ideas that involves medget modeling constructs, how it is presented and how flow of information be shown. Evaluation of the designs were done through a form of a survey. Conclusions drawn from the user survey is that, the most commonly used approach is still the most opted approach by the users. Also, information should be presented in the least possible way, but should be as descriptive as possible whether it is in a graphical form or not.