Counting Human Flow with Deep Neural Network

Abstract

Human flow counting has many applications in space management. This study applied channel state information (CSI) available in IEEE 802.11n networks to characterize the flow count. Raw inputs including mean, standard deviation and five-number summary were extracted from windowed CSI data. Due to the large number of raw inputs, stacked denoising autoencoders were used to extract hierarchical features from raw inputs and a final layer of softmax regression was used to model the flow counting problem. It is found that this deep neural network structure beats other popular classification algorithms including random forest, logistic regression, support vector machine and multilayer perceptron in predicting the flow count with attractive speed performance.