Application of wavelet threshold de-noising to sound travel-time estimation

The key to stored grain temperature measurement by acoustic method is to measure sound travel-time exactly. Grain is a highly absorbing acoustic medium, as a result, the sound signal received by the microphone far from the sound source has very low signal-to-noise ratio. In order to improve the sound travel-time estimation accuracy, a wavelet denoising-based cross correlation approach (in short CC_WTDN) is proposed. Soft threshold is used for its nice mathematical properties. Four threshold selection rules, that is, VisuShrink estimator, SUREShrink estimator, Heuristic estimator and Minimax estimator are tried. Simulation results based on a acoustic model of grain show that basic coross correlation method (BCC) can work well only when the sound wave path is not longer than 7.07m; while CC_WTDN can give good estimations even if the sound wave path is long as 10m. The four threshold rules tried give similar de-noising effects. The levels of wavelet decomposition have significant influence on de-noising effect.