Feature extraction and attribute recognition of particle light scattering signals based on wavelet scattering transform and long and short term memory network

Accurate identification and monitoring of particulate matter is an important means to solve the problem of atmospheric particulate pollution. Therefore, we propose a method of feature extraction and attribute recognition based on wavelet scattering and long and short term memory neural network. In this paper, the light scattering signals of particles with different attributes are collected by the experimental platform. Firstly, the EMD-ICA noise reduction model is employed to complete the noise reduction preprocessing of the particle light scattering signals. Then, the feature of scattering coefficient is extracted by wavelet scattering network, and the scattering feature matrix is constructed and input into the long and short term memory neural network for training. Finally, the probability classification of softmax layer is utilized recognize the attributes of different particles. Meanwhile, the proposed new method of particle property identification has a 98.83% correct classification rate, which shows that it has good performance to achieve particle property identification. Therefore, it provides a feasible basis for preventing air pollution and rapid particle identification.