Orientation Recognition of Human Postures and Activities based on IMU Devices

The goal of this study was to develop a model for human activity recognition that could distinguish between six human activities—standing, sedentary, walking, walking up and down stairs, and postural transitions between the standing and sedentary classes—regardless of how the smartphone was held in the wearer's pocket. Sensor readings can be represented in a consistent frame of reference by determining the orientation of the instrument using a complementary filter based on quaternions. Pitch/roll and tilt are two newly created features that measure the angle between the estimated upright orientation and the most recent average orientation of the IMU as well as the angle between the estimated upright orientation and the estimated average orientation of the IMU when walking is detected. Both of these features have been found to be helpful in classifying human activities. The success of these quaternion-derived features indicates that current methods for detecting human activity would benefit from transferring all measurements to the global frame of reference, where the feature values would be more uniform, particularly if the orientation of the IMU with respect to the body is not fixed.

کلیدواژه ها:

Human Activity ، IMU Sensors ، Classification ، Orientation ، Wearable Devices