Histogram Feature-Based Approach for Walking Distance Estimation Using a Waist-Mounted IMU

This paper presents a new method of walking distance estimation using an inertial measurement unit (IMU) placed on the user's waist. When the sensor is mounted on the waist, the walking steps can be easily detected. However, step length estimation is a challenging task. In this paper, a walking distance estimation method based on histogram features is proposed. The histogram features, including the uniform and logarithmic quantizations, are derived from each step segment of the acceleration norm data. Then, machine learning algorithms such as support vector regression (SVR), gaussian process regression (GPR), and linear regression (LR) are applied based on the histogram features to estimate the walking distance. Two experiments are conducted to evaluate the walking distance estimation accuracy, where test walking paths consist of a straight line corridor (80 m) and a rectangular path (about 1282 m). Experimental results show that the average absolute errors using the SVR model are 0.76% for straight line corridors and 1.14% for the rectangular paths. The proposed method is shown to outperform existing methods from comparison study.