Target Classification in Sparse Sampling Acoustic Sensor Networks using DTWC Algorithm

To extract as much accurate information as possible, especially in the case of a sparse sampling acoustic sensor network, the approach of time series can be effective. However, both problems of local time shifting and spatial variations should be solved to apply the time series analysis. This paper proposes the DTWC (DTW-Cosine) algorithm, as a time series manner, to solve the two problems and proves the performance through several experiments. We also considered acoustic variations, which can occur, by using data set mixed with various effects as input. Our experimental results show that the target classification rate of our algorithm not only outperforms the other time-warped similarity measure algorithms but it also has a robust performance over various volumes in combination with a smoothing technique. Since this proposed algorithm produces such a satisfactory result with sparse sampling data, it allows us to classify objects with relatively low overhead.