Hand Gesture Recognition Using Convolutional Neural Networks

This paper introduced a hand gesture recognition method based on convolutional neural networks (CNNs). The recognition scenario consisted in a three dimensional radar array to transmit and receive 24GHz continuous electromagnetic (EM) wave, and convert the scattered EM wave to the intermediate frequency (IF) signals. This paper used the the processed frequency spectrum as the input to the CNN. Then the CNN feature detection layer learned through data training, avoiding supervised feature extraction while learning implicitly from training data. It highlighted these features through convolution operating, pooling and a softmax function. Results showed that this system could achieve a high recognition accuracy rate higher than 96%.