This project proposes a solution for a pedestrian tracking using technologies that cannot be easily disrupted these days, based on orientation measurement components that are passively attached to a pedestrian’s shoe.
The system’s sensor that is used is a small and cheap component known as BOSH BNO-055, which contains 3 accelerometers, 3 gyroscopes, and 3 magnetometers. The tracking is done while the sensor is attached to a shoe and connected by a cable that transmits information to the laptop, where it is processed using location determination algorithms developed by us.
Unlike other projects that were done in the past, this system is attached to the shoes of the pedestrian, and as a result requires dealing with increased measurement noise.
After reviewing literature and past work in the field, developing a mathematical model, and understanding the capabilities and limitations of the sensor array when it is attached to a foot, we began with analyzing the components and the hardware system, a theoretical and mathematical analysis of the problem.
Then, we continued with the practical development of step detection.
As part of step event detection, this project deals with leg and body movements that may be distracting and causing false detections as step events. This was done by performing pre-processing part on the data, including sensor vibration noise filtering, axis coupling separation, and modeling of drifting in the sensor’s measurements.
In addition to general step detection, we have defined an adaptive model that is calibrated personally for the pedestrian. The creation of this model inserts robustness for changing pedestrians and various walking types, which allows us to meet the needs of many different users in the system.
After that, when the ability to recognize a step for the user was complete, we moved on to building the overall walking path of the pedestrian, which includes the development of a decision tree for classification of different types of movements and steps, adapting the algorithm for building a path in real time, and step size estimation both by empirical statistics and by theoretical research to find the right method for this project’s needs. As part of the project, the path mapping algorithm is robust to various walking types and turnings that are not sharp for being suitable for a more “natural” walk. As well as, adding a degree of freedom of movement in directions, twists, going up and down stairs, turning, curvy paths, walking backwards, and turning.
Eventually, we analyze our performances and examine their improvement compared to past projects.