Deep Learning-Based Gait Recognition Using Smartphones in the Wild (1811.00338v3)

Abstract: 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

This paper presents a paper on gait recognition via deep learning using smartphone-integrated inertial sensors under unconstrained conditions, referred to as "in the wild." Gait, as a biometric, offers advantages such as unobtrusiveness and resilience against concealment. Smartphones, which are widely accessible, contain such sensors, making them a practical medium for gait data acquisition.

The proposed method utilizes a hybrid deep-learning architecture combining Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) to effectively extract and model gait features from freely-collected inertial data. This approach is distinct from traditional methods, which often depend on structured and controlled data collection environments.

Key Insights and Results:

• Novel Architecture: A novel hybrid neural network combining CNNs with one-dimensional kernels and Long Short-Term Memory (LSTM) networks is introduced. This architecture efficiently extracts spatial and temporal features from six-dimensional time-series data (comprising accelerometer and gyroscope readings along x, y, and z axes).
• Data Handling and Experimentation: The paper evaluates the method on two datasets involving 118 subjects. The experiments focus on two core tasks: person identification and authentication via gait recognition. Results show identification and authentication accuracies exceeding 93.5% and 93.7%, respectively, demonstrating the method's effectiveness.
• Data Collection without Constraints: Unlike controlled environments, this method assumes the smartphone's inertial data capture occurs without restrictions regarding user location, time, or walking conditions. Fully Convolutional Neural Networks (FCNN) aid in segmenting the raw data into walking and non-walking sessions.

Implications and Future Directions:

The paper confirms the effectiveness of deep learning in processing unconstrained gait data, signifying an advancement in both theoretical understanding and practical applications of gait biometrics in everyday environments. The ability to unobtrusively and accurately authenticate individuals using ubiquitous mobile hardware presents significant potential for enhancing personal security applications, health monitoring, and secure access control systems.

Future research could expand on the presented architecture to incorporate multi-source data integration, adapt to varying gait patterns due to physical condition changes, or scale across larger, more diverse populations. Further exploration into real-time processing capabilities on mobile devices could bridge the gap between research environments and practical, everyday usability of gait-based biometric systems.