Abstract:  Gesture typing is a mode of input where the user draws a continuous path 

on a keyboard to interact with their device. Through proper data augmentation, a 

smaller dataset can be taken and given additional data to be used for training a 

supervised recognition model to recognize gestures. We propose a conditional GAN 

architecture to generate realistic keyboard gestures after training on a dataset of real 

user swipes. The model takes a gaussian noise vector and the straight-line path 

prototype for the target word and outputs a new gesture, using the straight-line path as 

a semantic map at multiple layers of the network. We show that this method can 

generate multiple different gestures for a word, without the need for reference gestures 

to generate, and even on words that have not been seen before. Through experiments 

on a dataset with unseen words, we evaluate our model against the minimum jerk 

model for generating gestures and show that our model outperforms it for generating 

realistic gestures.