Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with
or
.
Download ZIP
Browse files

Fixed typos, renamed lingering widget to activity classifier widget.

  • Loading branch information...
commit 4485111369315025882316bdfb86c1e75e6a5be0 1 parent aea2731
@jwf authored
Showing with 2 additions and 2 deletions.
  1. +2 −2 more-info.html
View
4 more-info.html
@@ -55,9 +55,9 @@
<p>To try out this plugin, you need to enable the sensor input plugin and the classifier output plugin. To enable the sensor input plugin, go to the <i>Settings</i>, then to <i>Input Plugins</i>, then click the check box next to <i>Sensor Logging Plugin</i>. In the <i>Sensor Speed</i> setting, choose <i>Fastest</i>. Then go back into the <i>Settings</i>, choose <i>Output Plugins</i>, and enable the <i>Simple Classifier Plugin</i>. The classifier is only active when the phone is in motion, and the Lingering parameters adjust how sensitive the classifier is. I've found that setting the <i>Lingering Power Threshold</i> to 100, and the <i>Lingering Power Window</i> to 25 work well. The <i>Manage Model Files</i> setting allows you to choose which models the classifier uses, and we'll get to that in a moment.</p>
<p>Now, return to the main <i>Humansense</i> application screen, and click <i>Start Service</i>. Two new buttons should appear at the bottom of the screen. <i>Build</i> is used to build new models (i.e., create new activities or people to be recognized). Clicking <i>Build</i> will cause the device to begin collecting samples from the accelerometers. To create a new activity model, click <i>Build</i>, put the phone in your pocket (locking the screen first so that no buttons are pressed accidentally while in your pocket), perform the activity for a short period of time (10 seconds or so), remove the phone from your pocket (unlock it), and click <i>Stop Building</i>. A new dialog will appear, showing you the collected accelerometer data. The purpose of this dialog is to trim off the initial and final few seconds, which contain noisy data from when you put the phone in your pocket and removed it. Select a nice section of the data by clicking and dragging on the data. A new dialog will appear allowing you to assign a name to this data. If you accidentally clicked on a bad portion of the data, click <i>Cancel</i> to return to the previous screen to re-select a segment. Once you have selected a good segment and assigned it a name that you will remember, click <i>OK</i>. This will return you to the main application screen. Repeat the process to create a few activities, or pass the phone to your friends and collect data from them walking around with the phone in their pocket.</p>
<div align="center"><img src="images/humansense-accel-trace.png" width="500" /></div>
- <p>Once you have collected a few activities, the next step is to perform one of the activities again, and have the phone guess which one you are performing. The first step is to select a set of candidate activities from which the classifier will choose. To do this, go into the <i>Settings</i>, into <i>Output Plugins</i>, and then select <i>Manage Model Files</i>. Here you will see a list of all of the models that you have created. Check off the models that you want to use for this session. At the moment, the application starts to stop performing well when more than 10 or so models are active. Once you have selected the models you would like to consider, click <i>Save</i> to save your model selection. Now go back to your phone's home screen (not the application home screen), and add the widget in which the results will be displayed. To do this, long-click on an empty portion of the home screen, select <i>Widget</i> and add the <i>Lingering Notification Widget</i>. The widget should appear on the home screen.</p>
+ <p>Once you have collected a few activities, the next step is to perform one of the activities again, and have the phone guess which one you are performing. The first step is to select a set of candidate activities from which the classifier will choose. To do this, go into the <i>Settings</i>, into <i>Output Plugins</i>, and then select <i>Manage Model Files</i>. Here you will see a list of all of the models that you have created. Check off the models that you want to use for this session. At the moment, the application starts to stop performing well when more than 10 or so models are active. Once you have selected the models you would like to consider, click <i>Save</i> to save your model selection. Now go back to your phone's home screen (not the application home screen), and add the widget in which the results will be displayed. To do this, long-click on an empty portion of the home screen, select <i>Widget</i> and add the <i>Activity Classifier</i> widget. The widget should appear on the home screen.</p>
<div align="center"><img src="images/humansense-model-selection.png" height="500" /></div>
- <p>Now, to start classifying, go back into the Humansense application, ensure that the service has been started, and then click the button on the bottom of the screen that says <i>Classify</i>. Now, if you go back to your home scren and look at the widget, you should notice that some text has appeared. The first two lines simply count the number of samples that are considered to be lingering (the phone is not moving), or otherwise. Below these two lines another set of lines should appear. There will be one line per activity, labeled according to the name that you chose when the model was built. Next to these activities will be the probability that the classifier has assigned to each activity. If you lock the phone and put it in your pocket, then perform one of the activities for a few seconds, when you unlock the phone the value next to the activity that you performed should be larger than the value for the other activities. As this is a work in progress, and part of an active research program, please let <a href="mailto:jordan.frank@cs.mcgill.ca">me</a> know how it works.</p>
+ <p>Now, to start classifying, go back into the Humansense application, ensure that the service has been started, and then click the button on the bottom of the screen that says <i>Classify</i>. Now, if you go back to your home screen and look at the widget, you should notice that some text has appeared. The first two lines simply count the number of samples that are considered to be lingering (the phone is not moving), or otherwise. Below these two lines another set of lines should appear. There will be one line per activity, labeled according to the name that you chose when the model was built. Next to these activities will be the probability that the classifier has assigned to each activity. If you lock the phone and put it in your pocket, then perform one of the activities for a few seconds, when you unlock the phone the value next to the activity that you performed should be larger than the value for the other activities. As this is a work in progress, and part of an active research program, please let <a href="mailto:jordan.frank@cs.mcgill.ca">me</a> know how it works.</p>
<div align="center"><img src="images/humansense-classifier-screenshot.png" height="500" /></div>
<h3>Location Recognition</h3>
Please sign in to comment.
Something went wrong with that request. Please try again.