Prediction and Analysis of ambient air particle pollutants using stationary wavelet transforms

Abstract

Ambient air particle pollutants (PM2.5 and PM10) are hazardous and play an important role in the air pollution focusing on human health and climate. Wavelet transforms extract local properties and information from a signal. In stationary wavelet transforms, the same number of samples as the input is maintained at every decomposition level and reconstructions result has lower error values and faster convergence compared to discrete wavelet transforms. The prediction work is carried out directly with the general linear predictor and wavelet predictor because linear predictor is a non-unique projection onto the wavelet domain. The approximation and detail represent average behaviour or trend and differential behaviour or changes of the signal respectively. The wavelet and statistical analysis for both original and extended signal are performed and discussed.