Characterizing a moving source in wireless sensor networks from the view of maximum likelihood

While trying to solve the problem of tracking unwelcome objects in wireless sensor networks (WSNs) based on acoustic signals, we notice that the sources' frequency components (f-components) and velocities play very important roles in improving the positioning results and for higher applications including source recognition. The information of a source like position, velocity, direction of moving and f-components are twisted in the data recorded at different sensors. This paper aims to derive all these characteristics based on the Doppler effect information. Since the outputs of Discrete Fourier Transformation of the collected data suffer from frequency leakage and Gaussian noise, we consider them as distribution functions on frequency domain. This is the novel idea to treat these functions as the likelihood distributions and calculate the hidden parameters during the attempt to maximize the join likelihood function. The method is described and then illustrated with experiments in the regard of estimation accuracy.