Sparse signal recovery for WIM measurements from undersampled data through compressed sensing

This study presents a method to recover the signal components critical for weigh-in-motion (WIM) measurements using compressed sensing. Through a comparative study, the wavelet basis 'bior2.4' is selected to sparsely represent the measured signals. Two methods, LASSO and Partial Inversion (Partlnv), are used to recover the critical pulses from numerically decimated signals. Research results show that, compared with the LASSO method, the Partlnv method improves the recovering accuracy of the pulse peaks by 70% on average and enhances the measurement reliability. This study offers a method of recovering the critical pulses for WIM and other types of measurements from undersampled signals. The proposed method enables the equipment designed for low-frequency measurements to achieve satisfactory measurements that requires a much higher sampling frequency. It has the potential to reduce the cost, energy, data storage, and data transmission requirements in practical implementations of undersampled WIM strain measurements and others. (C) 2019 Elsevier Ltd. All rights reserved.