Detail publikace

Artificial Neural Networks in an Inertial Measurement Unit

Originální název

Artificial Neural Networks in an Inertial Measurement Unit

Anglický název

Artificial Neural Networks in an Inertial Measurement Unit

Jazyk

en

Originální abstrakt

This paper presents an effective method combining classic data processing using a simple MEMS inertial measurement unit (IMU) and an artificial neural network (AAN) to achieve more accurate pedestrian positioning. Generally, this application based on a standard IMU without support from another system, such as satellite navigation, is characterized by poorly estimating position and orientation, wherein the positioning error grows over time. The proposed approach uses an artificial neural network, which is designed to determine the status of "what is happening" with the body of the IMU. Two possible statuses are considered. The first of these is the fact that the IMU is static, regardless of its orientation, and the second state is a man walking with an IMU placed on his body. In principal, further statuses can be added to the classification results from the ANN, e.g. jogging, driving, shaking, spinning, flying, falling etc. This paper not only presents the theoretical but also a series of experiments. It has been demonstrated that the proposed approach improves personal tracking accuracy by more than ten times compared to the application of an unaided IMU.

Anglický abstrakt

This paper presents an effective method combining classic data processing using a simple MEMS inertial measurement unit (IMU) and an artificial neural network (AAN) to achieve more accurate pedestrian positioning. Generally, this application based on a standard IMU without support from another system, such as satellite navigation, is characterized by poorly estimating position and orientation, wherein the positioning error grows over time. The proposed approach uses an artificial neural network, which is designed to determine the status of "what is happening" with the body of the IMU. Two possible statuses are considered. The first of these is the fact that the IMU is static, regardless of its orientation, and the second state is a man walking with an IMU placed on his body. In principal, further statuses can be added to the classification results from the ANN, e.g. jogging, driving, shaking, spinning, flying, falling etc. This paper not only presents the theoretical but also a series of experiments. It has been demonstrated that the proposed approach improves personal tracking accuracy by more than ten times compared to the application of an unaided IMU.

BibTex


@inproceedings{BUT125540,
  author="Lenka {Tejmlová} and Jiří {Šebesta} and Petr {Zelina}",
  title="Artificial Neural Networks in an Inertial Measurement Unit",
  annote="This paper presents an effective method combining classic data processing using a simple MEMS inertial measurement unit (IMU) and an artificial neural network (AAN) to achieve more accurate pedestrian positioning. Generally, this application based on a standard IMU without support from another system, such as satellite navigation, is characterized by poorly estimating position and orientation, wherein the positioning error grows over time. 
The proposed approach uses an artificial neural network, which is designed to determine the status of "what is happening" with the body of the IMU. Two possible statuses are considered. The first of these is the fact that the IMU is static, regardless of its orientation, and the second state is a man walking with an IMU placed on his body. In principal, further statuses can be added to the classification results from the ANN, e.g. jogging, driving, shaking, spinning, flying, falling etc. 
This paper not only presents the theoretical but also a series of experiments. It has been demonstrated that the proposed approach improves personal tracking accuracy by more than ten times compared to the application of an unaided IMU.",
  booktitle="Proceedings of the 26th International Conference RADIOELEKTRONIKA 2016",
  chapter="125540",
  howpublished="online",
  year="2016",
  month="april",
  pages="176--180",
  type="conference paper"
}