Vehicle Classification in Acoustic Sensor Networks Based on Hybrid Dictionary Learning

In this paper, we consider a single sensor classification problem, focusing on classifying the types of the moving vehicles. To improve the classification accuracy with low-time complexity in complex scenes, the acoustic sensor data sets were captured to measure the physical events and a novel hybrid dictionary learning method for vehicle classification is proposed. The efficient hybrid dictionary is constructed by a synthesis dictionary and an analysis dictionary, in which the analysis dictionary is trained to generate discriminative codes and the synthesis dictionary is trained to achieve class-specific discriminative reconstruction. To further analyze the performance of this proposed method, we extend our approaches to deal with sparse environmental noise. Extensive experiments are conducted on real data sets and the results demonstrate that our proposed approach yields higher classification accuracy with relatively lower time complexity compared to SVM, SRC, KSRC, LC-KSVD and FDDL algorithms.