Advanced seismic signal processing techniques for detection and bearing estimation of footsteps and vehicular signals

• Main Author: Divya Venkatraman
• Other Authors: Andy W. H. Khong
• Format: Theses and Dissertations
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: https://hdl.handle.net/10356/69244
• Institution: Nanyang Technological University

This thesis is aimed towards the development of algorithms for improved detection and bearing estimation of human activities in urban surveillance systems. Human footsteps and vehicles are used as sources and seismic signals, generated from these sources, are recorded using a (buried) tri-axial geophone. The performance of existing energy-based metrics deteriorate signi cantly with increasing source-sensor distance due to poor signal-to- noise ratio. The three-dimensional (3D) trajectory traced by the tri-axial signal, acquired from near-fi eld seismic sources, is elliptically polarized. This thesis takes as its point of departure the direction attribute of the elliptically polarized 3D trajectory to detect and estimate the direction of arrival of surface seismic sources. The smoothness of the 3D trajectory is used as a feature and instantaneous metrics using consecutive snapshots of the normalized trajectory are de ned. Furthermore, the 3D trajectory is analyzed using two di fferent representations of the hypercomplex quaternions. Using the vector-quaternion representation a quaternion generalized Gaussian distribution (QGGD) is derived. The QGGD is parameterized by a shape parameter and a quaternion augmented covariance matrix to capture non-Gaussianity and inter-channel correlation, respectively, of the tri-axial signals. The orientation-quaternion representation is used to parameterize the instantaneous orbital phase of the trajectory. The concentration and direction of the orbital phase are used to detect the presence of elliptical polarization and to estimate the bearing of seismic surface sources, respectively. The proposed trajectory-based metrics were found to improve the detection range of human footstep signals by 5 m to 7 m and vehicular signals by 25 m to 100 m, compared with existing energy-based detection metrics. The orientation-quaternion representation of the 3D trajectory enhanced bearing estimation accuracy, in addition to addressing the 90 ambiguity that occurred in existing methods.