Gesture recognition is a very essential technology for many wearable devices.While previous algorithms are mostly based on statistical methods including thehidden Markov model, we develop two dynamic hand gesture recognition techniquesusing low complexity recurrent neural network (RNN) algorithms. One is based onvideo signal and employs a combined structure of a convolutional neural network(CNN) and an RNN. The other uses accelerometer data and only requires an RNN.Fixed-point optimization that quantizes most of the weights into two bits isconducted to optimize the amount of memory size for weight storage and reducethe power consumption in hardware and software based implementations.
展开▼
