Robust Keyword Spotting via Recycle-Pooling for Mobile Game

We present an effective method to solve a small-footprint keyword spotting (KWS) task via deep neural network for mobile game. Our goal is to improve the accuracy of KWS in various environments. To this end, we propose a new neural network layer named recycle-pooling. Extensive experiments indicate that our recycle-pooling based convolutional neural network (RP-CNN) indeed improves the performance of KWS in both clean and noisy data for mobile game. We will perform live demonstration of RP-CNN based KWS integrated into a full-sized, production-quality mobile game A3: Still Alive, which is one of the major games from Netmarble this year and will be available on market soon.