Footstep detection using Laplacian distribution

Human activity detection becomes important aspect in zone monitoring and protection systems in military scenarios. This work is focused on comparing the types of detectors and thereby suggesting the best possible detector for multiple sample signals of footsteps. Multiple features were extracted and analysed. A Log-Likelihood Ratio Test detector has been designed, assuming that the DFT Coefficients follow either a Laplacian or a Gaussian distribution. Among Laplacian based and Gaussian based detectors, Laplacian has proved to perform better in samples with lower SNR values, while both Laplacian and Gaussian detectors perform equally well in the higher SNR signals.