Compared to other biometrics, gait is difficult to conceal and has the advantage of being unobtrusive. Inertial sensors, such as accelerometers and gyroscopes, are often used to capture gait dynamics. These inertial sensors are commonly integrated into smartphones and are widely used by the average person, which makes gait data convenient and inexpensive to collect. In this paper, we study gait recognition using smartphones in the wild. In contrast to traditional methods, which often require a person to walk along a specified road and/or at a normal walking speed, the proposed method collects inertial gait data under unconstrained conditions without knowing when, where, and how the user walks. To obtain good person identification and authentication performance, deep-learning techniques are presented to learn and model the gait biometrics based on walking data. Specifically, a hybrid deep neural network is proposed for robust gait feature representation, where features in the space and time domains are successively abstracted by a convolutional neural network and a recurrent neural network. In the experiments, two datasets collected by smartphones for a total of 118 subjects are used for evaluations. The experiments show that the proposed method achieves higher than 93.5% and 93.7% accuracies in person identification and authentication, respectively.

Deep Learning-Based Gait Recognition Using Smartphones in the Wild

Multi-sensor information fusion based on machine learning for real applications in human activity recognition: State-of-the-art and research challenges

Adaptive gait detection based on foot-mounted inertial sensors and multi-sensor fusion

Pedestrian Dead Reckoning (PDR) is the process of calculating one's current location by using the previously known position, and advancing that position over time using established or estimated speeds and trajectories (or alternatively stride lengths and directions). PDR plays an important role in modern life, including tracking locations of people and objects whenever GPS is not available. Self-contained PDR systems do not require an infrastructure, thus they can be used for rapid deployment in situations such as search and rescue, disaster relief or medical emergencies. Wearable sensors are often applied in self-contained PDR, but implementation varies in terms of the number, type and location of sensors used. Many algorithms are designed for PDR in order to reduce the error or drift of the final estimate, with various levels of success. There is a lack of comparison between these different methods and this systematic review of PDR for wearable devices provides a comprehensive overview that can inform further design optimizations. The aim of this article is to assess the quality of all available PDR literature with a focus on wearable sensors. It provides an outline of the state-of-the-art in the field by comparing the accuracy of different sensor layouts and algorithms. Further directions of research are suggested based on these results. This study also highlights the need for more standardised and robust assessment protocols to capture real-world tracking performance of PDR methods.

/pdf/pedestrian-dead-reckoning-with-wearable-sensors-a-systematic-50mxtz56gr.pdf

Pedestrian Dead Reckoning With Wearable Sensors: A Systematic Review

Foot-mounted inertial navigation is an important issue in areas such as pedestrian localization, gait analysis, and sport training. However, low-cost inertial sensors suffer from several errors that make the navigation results less convincing. In this paper, a multi-sensor approach with one sensor on each foot is presented to reduce the system heading drift. Through dual-gait analysis, gait parameters between two feet are employed to make the non-collocated and uncoupled subsystems be related to each other. A step length estimator based on an inverted pendulum model is developed to derive a relative position vector between the two foot-mounted sensors rather than a distance scalar. A Kalman-type filter with one time update and two measurement updates is developed to fuse the velocity and position observations at foot and person levels, respectively. Experiments were conducted by four healthy subjects, and experimental results show that the proposed sensor fusion method can effectively reduce the heading drift of inertial navigation and make the captured dual-foot motion closer to its actual process.

Heading Drift Reduction for Foot-Mounted Inertial Navigation System via Multi-Sensor Fusion and Dual-Gait Analysis

Indoor pedestrian dead reckoning (PDR) extends the location-based service (LBS) to environments, where GPS or beacon signals are degraded or unavailable. A practical PDR system should consider the absence of any infrastructure or prior knowledge of the environment. This paper presents a PDR system based on a pocket-worn smartphone, which tracks a person's location through dead reckoning calculation by using the sensors embedded in smartphones. However, a smartphone-based PDR system faces various challenges, especially the heading drift due to gyroscope bias. In this paper, the gradient descent algorithm (GDA) is improved to reduce the heading drift, by fusing inertial data with only a fraction of magnetometer data that are accurate and usable. There is an 80% probability that the heading error is reduced to less than 4 degrees. Besides, a stride detection method is developed based on thigh angles, and then, a stride length estimation method is implemented in a complementary way. The experiments were conducted along with three types of reference paths, and the experimental results demonstrate that the average position errors along the three paths are 1.62%, 1.00%, and 0.92%, respectively. Despite the inherent sensor noise and complex human locomotion, the smartphone-based PDR system has great potential in pedestrian tracking.

Luyao Zhang

Papers

Heading Drift Reduction for Foot-Mounted Inertial Navigation System via Multi-Sensor Fusion and Dual-Gait Analysis

Pedestrian Dead Reckoning Using Pocket-Worn Smartphone