Activity recognition for indoor movement and estimation of travelled path

Nowadays, human activity recognition with wearables is an interesting field, but the accurate assessment of activity remains a challenging problem because of the different behaviour in which the activities are performed of different persons. In this paper machine learning with support vector machines in combination with an Inertial Measurement Unit (IMU) is used to determine the different kinds of indoor movement, like standing, walking or climbing ascending and descending stairways. And because indoor positioning is still an interesting theme, the detected human activity in combination with the accelerometer data of the IMU can be used for determination of the travelled path to help to locate the person indoor or to support indoor positioning systems. To limit the costs of the device, the approach is made to develop a system with the use of as few data as possible to reduce the hardware costs and the size of the device.