Source separation of multichannel acoustic signals measured from living trees
Taigang He, Maria C. Hansson, Joakim Axmon · 2002
The method in this paper is applied to vibration analysis of the tree trunk. The measured vibrations are believed to provide information about wave coupling and propagation, which may contribute to the detection of decay inside trees. A new signal model is presented and a source separation method using independent component analysis is applied. The aim of the study is to separate the multiple measured signals into independent sources for further analysis. The performance of the method used is investigated using simulated data and real data measured from living trees. The results of simulations validate the proposed method, while the applications justify the signal model, showing that the real multichannel data can be successfully separated into sources similar to harmonic oscillations.