Harukaze

Initially, when I learnt that I was gonna move from developing an application on a Winform to developing it on a Webform, I thought it’ll be a pretty quick and easy transition. After all, they’re both in C# and part of the .NET framework, eh?

And as always, my eagerness to assume bites me in the ass again. Application Life Cycle vs Web Form Life Cycle, Instanced Variables vs Session Variables, Callbacks vs Postbacks…and far too many more differences for my liking.

Nevertheless, it’s gonna be another interesting albeit bumpy ride in my journey through game development. As the wise sages in the mountaintops spake thusly, “What don’t kill you makes you stronger.”

Gesture recognition seems to be a pretty interesting step up from the traditional forms of human-computer-interaction via keyboard, mouse or controllers. While gesture recognition technology started off(to my knowledge) as being used in personal communication with PDAs, game developers have caught on to the possibilities of implementing it in games. The first game I came across that made use of that was Black & White, where the user plays a god that could cast ’spells’ based on the runes drawn on the screen. Although it did not constitute a major portion of the gameplay, it was still an interesting concept at that time.

I think game developers started looking into this more seriously with the launch of the Nintendo DS. Most games developed for the handheld console used the stylus more as a clicker, but there were a few which set themselves apart from the rest by implementing gestures as well, like Okamiden and Castlevania Dawn of Sorrow.

With traditional input, you will always get a well-defined set of outcomes from the user, or at least, a set of outcomes that a normal person can visualize and categorize. However, with the addition of user gestures, the possibilities increase exponentially, even within a small area of interaction. The trick is to find a way to turn all that seemingly random data into something recognizable that can be categorized easily.

And that is going to be the basis of my research for now.

From what I’ve gleaned from online sources so far, there seems to be a few popular solutions, from translating the input into histograms and accelerometer data, to machine learning with neural networks.

However, for me, I’m going to try to do it a simpler way, with vector averaging. I feel that it’ll be something a bit more easier to understand and implement, that even novice game programmers can get the hang of too. Efficiency-wise, probably not that great, but I’ll leave that as my next topic of research.