Americo G. Woolard

Towards indoor localization of pedestrians via smart building vibration sensing

Indoor localization by means of GNSS or a cellular-based method is known to be difficult. Potentially, other wireless technologies could address the technical requirements, but they usually imply the end user must carry a device compatible with this additional technology too. In this paper we investigate the feasibility of collecting vibration sensor readings within a building to locate pedestrians by their footsteps. Vibration propagation in buildings is markedly different than radio wave propagation in free space, thus prompting one to question the suitability of conventional positioning algorithms for this task. We presents the results of experiments conducted with actual measurements from an instrumented, smart building. We expect such buildings to become more prevalent in the future thanks to the technical advances and cost reductions provided by the Internet-of-Things (IoT). The promising initial findings indicate that time-difference-of-arrival, within a limited spatial extent, could be a viable localization technique, and these results encourage further research into vibration-based indoor localization.

Vibration Event Localization in an Instrumented Building

In this paper, we present the preliminary results of an indoor location estimation campaign using real data collected from vibration sensors mounted throughout an instrumented smart building. The Virginia Tech Smart Infrastructure Laboratory house a unique testbed featuring a fully instrumented operational building with over 240 accelerometers permanently mounted to the steel structure. It is expected that in the future, more and more buildings will be constructed with sensors scattered about their infrastructures, in no small part due to the envisioned promises of such systems which include improved energy efficiency, health and safety monitoring, stronger security, improved construction practices, and improved earthquake resistance. One of the most promising uses of this smart infrastructure is for indoor localization, a scenario in which traditional radio-frequency based techniques often suffer. The detection and localization of indoor seismic events has many potential applications, including that of aiding in meeting indoor positioning requirements recently proposed by the FCC and expected to become law in the near future. The promising initial results of a simplistic time-difference-of-arrival based localization system presented in this paper motivate further study into the use of vibration data for indoor localization.

Gender Classification of Walkers via Underfloor Accelerometer Measurements

The ability to classify the gender of occupants in a building has far-reaching applications including security and retail sales. The authors demonstrate the success of machine learning techniques for gender classification. High-sensitivity accelerometers mounted noninvasively beneath an actual building floor provide the input for these machine learning methods. While other approaches using gait measurements, such as vision systems and wearable sensors, provide the potential for gender classification, they each face limitations. These limitations include an invasion of privacy, occupant compliance, required line of sight, and/or high sensor density. Underfloor mounted accelerometers overcome these limitations. The authors utilize the highly-instrumented Goodwin Hall smart building on the Virginia Tech campus to measure vibrations of the walking surface caused by walkers. In this paper, the gait of 15 individual walkers was recorded as they, alone, walked down the instrumented hallway. Fourteen accelerometers, mounted underneath the walking surface, recorded walking trials with the placement of the sensors unknown to the walker. This paper studies bagged decision trees, boosted decision trees, support vector machines, and neural networks as the machine learning techniques for their ability to classify gender. A tenfold-cross-validation method is used to comment on the validity of the algorithms ability to generalize to new walkers. This paper demonstrates that a gender classification accuracy of 88% is achievable using the underfloor vibration data from the Virginia Tech Goodwin Hall by using decision tree approaches.

Indoor positioning from vibration localization in smart buildings

Indoor localization by means of a global navigation satellite system (GNSS) remains a difficult problem due to GNSS signal impairments created by the buildings structure. This problem prompted the research community to devise many alternative techniques. Unfortunately, in order to locate persons indoors, these alternatives often require each person to carry some device to facilitate the localization process. This paper investigates the naturally-generated vibration signals from a persons footsteps as a potential source of information for indoor localization. Instrumenting a building with vibration sensors is a mature technology, but, historically, the role of the technology was measuring a buildings response to external events (e.g., earthquakes), not for measuring occupant-generated vibrations. Some prior work studied outdoor detection of footsteps near borders or within restricted areas, but that environment and the localization objectives differed sufficiently from the scope of this research to limit the relevance of prior results. This paper reports on measurements from an instrumented, public building and examines viability of conventional localization algorithms for locating persons moving within a building. Noting the sub-optimum performance of these algorithms in this localization task, this paper proposes an extension to existing techniques to accommodate signal distortions encountered by vibrations in building structures.

Detection and Identification of Firearms Upon Discharge Using Floor-Based Accelerometers

Vibration monitoring and analysis techniques have significant potential to improve security and threat detection in the built environment. The cornerstone of the Virginia Tech Smart Infrastructure Laboratory (VTSIL) is the highly instrumented Goodwin Hall on the VT campus. This 5-story classroom and laboratory building is instrumented with over 200 accelerometers hard-wired throughout its 160,000 square feet, providing a platform for research and education in structural health monitoring, dynamic model validation, and occupancy studies, among other smart building applications. One of the major research goals for VTSIL is to utilize vibration data to develop advanced security strategies, including threat detection, identification, and localization. Toward realizing this goal, a mobile cement and I-beam platform was built and instrumented with accelerometers. This test-bed recorded vibration signatures during the event of a person discharging a firearm while standing atop the platform. This paper includes initial results that demonstrate there are detectable differences in sensor measurements between a handgun, rifle, and shotgun. Initial analysis of this vibration data using the singular value decomposition demonstrates that one can deduce the type of firearm discharged regardless of differences in the shooter (male, female, weight, etc.), thus justifying the pursuit of advanced vibration-based threat detection and identification systems.

Assessment of Large Error Time-Differences for Localization in a Plate Simulation

The advent of structural building instrumentation invites research into novel applications of such systems. Previous research has shown that the propagative nature of ground impacts on the floor of a building can be assimilated to a thin plate. This research presents results for time-difference of arrival (TDOA) and cross-correlation methods used for source localization in a sparsely instrumented plate Finite Element Model (FEM), where acceleration data is used. The overall accuracy is evaluated for various wave speeds with two different configurations of sensor positions, and the consistency of the perceived wave speeds are assessed by the coefficient of variance. The accuracy of localization is considered, with emphasis on the sign of arrival (SOA), as a method to provide directional inference. Peak-difference and cross-correlation techniques are found to be significantly erroneous, while the SOAs are consistently accurate, with less than 10 % of 126 SOAs being reported incorrectly for all cases. Repositioning the sensors closer to the boundary increased the errors for all methods.

In-field implementation of impedance-based structural health monitoring for insulated rail joints

Track defects are a major safety concern for the railroad industry. Among different track components, insulated rail joints, which are widely used for signaling purposes, are considered a weak link in the railroad track. Several joint-related defects have been identified by the railroad community, including rail wear, torque loss, and joint bar breakage. Current track inspection techniques rely on manual and visual inspection or on specially equipped testing carts, which are costly, timeconsuming, traffic disturbing, and prone to human error. To overcome the aforementioned limitations, the feasibility of utilizing impedance-based structural health monitoring for insulated rail joints is investigated in this work. For this purpose, an insulated joint, provided by Koppers Inc., is instrumented with piezoelectric transducers and assembled with 136 AREA rail plugs. The instrumented joint is then installed and tested at the Facility for Accelerated Service Testing, Transportation Technology Center Inc. The effects of environmental and operating conditions on the measured impedance signatures are investigated through a set of experiments conducted at different temperatures and loading conditions. The capabilities of impedance-based SHM to detect several joint-related damage types are also studied by introducing reversible mechanical defects to different joint components.

Evaluation of a new source localization method in a simulated dispersive plate

The problem of estimating the location of an impact force in a dispersive medium is complicated given the dispersion-related distortion of the generated traveling wave. The problem cannot be solved, with reasonable accuracy, using conventional time difference of arrival (TDOA) techniques. A building floor is an example of a dispersive medium that is being loaded by occupant footsteps. If more accurate localization algorithms are obtained, then they can be used to localize and track occupants in a building using floor vibration sensors measuring the footstep-induced traveling waves. This paper presents the evaluation of a new localization approach, in a simulated aluminum plate (dispersive waveguide), using a network of sensors measuring the plates vibration. Average signal power is calculated for all the sensors over a fixed time period, and then used to generate a location estimate. Two different location estimation solutions are presented and compared; a constrained least squares solution (CLS), and a non-linear root finding solution generated using the Levenberg-Marquardt (LM) algorithm. A finite element (FE) thin plate model is used as a testbed to evaluate the performance of the developed localization algorithm by estimating the location of virtual hammer impacts acting on the plate. The results encourage further future development.

Design and development of a prototype platform for gait analysis

The field of event classification and localization in building environments using accelerometers has grown significantly due to its implications for energy, security, and emergency protocols. Virginia Tech’s Goodwin Hall (VT-GH) provides a robust testbed for such work, but a reduced scale testbed could provide significant benefits by allowing algorithm development to occur in a simplified environment. Environments such as VT-GH have high human traffic that contributes external noise disrupting test signals. This paper presents a design solution through the development of an isolated platform for data collection, portable demonstrations, and the development of localization and classification algorithms. The platform’s success was quantified by the resulting transmissibility of external excitation sources, demonstrating the capabilities of the platform to isolate external disturbances while preserving gait information. This platform demonstrates the collection of high-quality gait information in otherwise noisy environments for data collection or demonstration purposes.

Classification of event location using matched filters via on-floor accelerometers

Recent years have shown prolific advancements in smart infrastructures, allowing buildings of the modern world to interact with their occupants. One of the sought-after attributes of smart buildings is the ability to provide unobtrusive, indoor localization of occupants. The ability to locate occupants indoors can provide a broad range of benefits in areas such as security, emergency response, and resource management. Recent research has shown promising results in occupant building localization, although there is still significant room for improvement. This study presents a passive, small-scale localization system using accelerometers placed around the edges of a small area in an active building environment. The area is discretized into a grid of small squares, and vibration measurements are processed using a pattern matching approach that estimates the location of the source. Vibration measurements are produced with ball-drops, hammer-strikes, and footsteps as the sources of the floor excitation. The developed approach uses matched filters based on a reference data set, and the location is classified using a nearest-neighbor search. This approach detects the appropriate location of impact-like sources i.e. the ball-drops and hammer-strikes with a 100% accuracy. However, this accuracy reduces to 56% for footsteps, with the average localization results being within 0.6 m (α = 0.05) from the true source location. While requiring a reference data set can make this method difficult to implement on a large scale, it may be used to provide accurate localization abilities in areas where training data is readily obtainable. This exploratory work seeks to examine the feasibility of the matched filter and nearest neighbor search approach for footstep and event localization in a small, instrumented area within a multi-story building.