Dynamic multidimensional sensor data acquisition with adaptive Kalman filtering

Abstract. Traditional mobile sensing systems often experience a decline in data acquisition accuracy in dynamic environments due to the use of fixed-parameter Kalman filters, which lack adaptability to changes in motion states and sensor noise. To address this limitation, this paper proposes a gait-aware adaptive Kalman filtering method that dynamically adjusts filter parameters based on real-time gait frequency analysis. This method enables autonomous real-time data collection of multi-dimensional data from multiple sensors embedded in smartphones. By employing adaptive covariance optimization, the method enhances the system's ability to handle different motion patterns, ensuring the accuracy and robustness of data collection. Experimental results demonstrate that this method improves data quality in diverse motion scenarios, providing a reliable foundation for multi-modal sensing and intelligent mobile sensing applications.