Reading 10

Visual Similarity of Pen Gestures

Citation:

Long Jr, A. Chris, et al. "Visual similarity of pen gestures." Proceedings of the SIGCHI conference on Human Factors in Computing Systems. ACM, 2000.

Publication Link: http://dl.acm.org/citation.cfm?id=332458

Summary:

Multidimensional scaling is used for reducing the dimensionality so that the matrices can visualized better for similarity. Regression analysis is done to find out which dimension represents what. 22 features were used to identify gestures, 11 of which were inherited from Rubine. 

This paper suggests that aspect is much more powerful than length and width of bounding box diagonal. The impact it creates is much higher. Also, the length and arc of shape have much lower impact. The log of aspect creates more sense than log because the values are normalized and easy to visualize for differences. 

Density metric is used for differentiating between coloring and scribbling. The curviness is measured with a threshold angle of 19 degree, which is got by summing small difference in angles.

Discussion:

I found the observation that log creates a better perception very intuitive. It certainly eases the comparison between two similar looking figures. Also, the selection of 19 degrees for the value f curviness could be explained a little bit more.

Reading 9

Citation:

Gesture Recognition Overview. 

No Link available.

Summary:

This chapter presents all the features used for linear classification in Rubine. The sampling rates are higher nowadays and because of that we might get points which give us the same location or time stamp. Thus, the corresponding divide by zero error needs to be handled by pre-processing. This may be either removing the first point or the second point or just altering the time stamp we have. The features like rotations, stroke lengths and the angle scenarios are used for identifying particular aspect of the gestures.

Long and Quill have used 11 of the 13 features of Rubine(removing the time features), but the performance improvement is not significant.

Discussion:

The chapter requires refinement in terms of explanation of Rotational measure and arctan discontinuity that it just goes over very rapidly. The way open questions are posed in the chapter is really thought provoking.

Reading 8

Specifying gestures by example

Citation:

Rubine, Dean. Specifying gestures by example. Vol. 25. No. 4. ACM, 1991.

Link: http://dl.acm.org/citation.cfm?id=122753

Summary:

Paper introduces an application called GRANDMA which is a toolkit for rapidly adding gestures to direct manipulation interfaces. Another application GDP, which is a gesture based application is also introduced in this paper.

GDP is also a single stroke recognizer, while it can be a pain in identification but it also avoids segmentation problems. The input gesture is classified into one of the C classes of gestures. Then, features are extracted from template and matched using a linear classifier. The paper also describes the Rubine's 13 features.

Discussion:

This paper could have used and tried more features along with Rubine features. The input technique provided by this paper is very promising. In my opinion, the major limitation is the identification of only single stroke gestures.

Reading 7

Citation:

$1 recognizer for user interface prototypes

Wobbrock, Jacob O., Andrew D. Wilson, and Yang Li. "Gestures without libraries, toolkits or training: a $1 recognizer for user interface prototypes."Proceedings of the 20th annual ACM symposium on User interface software and technology. ACM, 2007.

Publication Link: http://dl.acm.org/citation.cfm?id=1294238

Summary:

1$ recognizer is an easy to use, and very cheap gesture recognizer and is very efficient for template matching. 16 different types of gestures which can recognized by the recognizer have been discussed in the paper. 

The paper matches previously recorded template points to new candidate points generated by the gesture at that time. 1$ recognizer has implemented algorithm in four steps: Resample the point path, Rotate once based on indicative angle, Scale and translate and lastly finding the optimal angle for best score. It converts the minimum path to a particular score which is further used for recognition. It is rotation invariant.  Both sides arrows are really different and thus path being identical is very important. 

The major limitation is that it is a  geometric template matcher, and thus candidate strokes are compared to previously stored templates. Also, it cannot distinguish between gestures whose identities depend on specific orientations. Also, it has to be single stroke for it to be recognized.

Discussion:
There should be some more test cases presented for more number of templates to give a better idea to the reader. Just three templates make the test data too small I believe. The idea of looking at centroid of gesture and not any other random point is really impressive.

Reading 6

Paper:

Link: Using Entropy to Distinguish Shape Versus Text in Hand-Drawn Diagrams

Citation:

Bhat, Akshay, and Tracy Hammond. "Using Entropy to Distinguish Shape Versus Text in Hand-Drawn Diagrams." IJCAI. Vol. 9. 2009.

Summary:

Distinguishing between text and shape is a critical task. The approach followed for it in this paper is finding the entropy of each thing and generally the entropy of text is much higher. This is because entropy measure is generally the measure of randomness of an information source. Also, handwritten text is more difficult to describe than common shapes. A zero order entropy is used.

In the entropy model used, each symbol corresponds to a range of angles. The first step in the classification process is to group the strokes. This step uses using the intuition that strokes that are part of the same entity should be close to each other and drawn within a small amount of time. After strokes are grouped, strokes are re-sampled to smooth and sub-stroke angles so that all points are equidistant from each other. Probability estimate is calculated by dividing the total number of occurrence of a symbol the total number of points. Then, all the probabilities within a stroke group are added.

Based on a threshold calculated from training dataset, the input stroke group is classified as either shape, text, or unclassified. Confidence is measured and modeled by using arctan function.

Discussion:

The paper is written in a very easy to read and understand way. The accuracy of the system with maximum accuracy mode is 95% and with more classification mode is 92%, which are pretty impressive results. I think it should be tested with more misclassification symbols also other than a resistor or a dot for better results. Also, some more inputs and results for higher order entropy models would have explained the situation a lot better.

Reading 5

Citation

 L. Kara, T. Stahovich. An Image-Based, Trainable Symbol Recognizer for Hand-Drawn Sketches. Computers & Graphics 29.4 (2005): 501-517.

 Link: http://www.sciencedirect.com/science/article/pii/S0097849305000853

Summary:

A trainable hand drawn symbol recognizer is described. The approach is useful for sketchy inputs as it is based on binary templates. This paper is designed around four different similarity methods to enhance recognition accuracy. These are Hausdorff distance, Modified Hausdorff distance, Tanimoto coefficient and Yule coefficient. During matching, distance maps are used as look-up tables for the closest distances. Polar coordinate transformation technique is used for rotation invariant recognition. Next step is examining remaining definitions in screen coordinates.

Discussion:

The idea of use of polar coordinate transformation for rotation invariant recognition is very impressive. The accuracy achieved, which is 95% is really commendable. The sensitivity to non-uniform scaling portrays the importance of topology over shape.

Reading 4

Citation:

Davis, Richard C., Brien Colwell, and James A. Landay. "K-sketch: a'kinetic'sketch pad for novice animators." Proceedings of the SIGCHI Conference on Human Factors in Computing Systems. ACM, 2008.

Link: http://dl.acm.org/citation.cfm?id=1357122

Summary:

K-sketch is a general purpose 2D animation sketching system. It is able to provide quick and powerful animations meant to be used by novices. Then, a definition of an optimization problem was done to maximize the number of features supported. At the same time, it minimized the number of steps needed to complete each animation feature, thus making it optimized. Evaluations comparing with simple tools such as Powerpoint and Flash were done. This implies the ease of use of the system.

Discussion:

The ease of use makes K-sketch a very optimal choice of system to use. The fact that ease of use of User interface could be compared with Powerpoint really amazed me. Also, considering animation as a step wise procedure, it shows the main purpose of the system is Novice users and their ease of use.