Academia.eduAcademia.edu

Figure 11 – uploaded by Stevens Osz

See full PDFdownloadDownload figure
FIGURE 11: Roll values. (a) Accelerometer. (b) Gyroscope. (c) Proposed method. (PD) and the Samsung health android application ap- proach. The error analysis shows that the Samsung health android application has very poor performance than other approaches and it is not suitable for PDR localization. Dekh td wenden) wm en ndiiinean). «LF «ndlLandetaesn “nn chenanes £5 awed uate the performance of proposed heading estimation al- gorithm. It consist of four direction paths. The proposed heading estimation algorithm uses the sensor fusion technique for estimating the direction. We compare the proposed heading estimation with azimuth based and magnetometer based approaches. The heading from all proach. The error analysis shows that the Samsung Path1 rectangular motion of pedestrian is chosen to eval- 2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. Sec http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Figure 11 Roll values. (a) Accelerometer. (b) Gyroscope. (c) Proposed method. (PD) and the Samsung health android application ap- proach. The error analysis shows that the Samsung health android application has very poor performance than other approaches and it is not suitable for PDR localization. Dekh td wenden) wm en ndiiinean). «LF «ndlLandetaesn “nn chenanes £5 awed uate the performance of proposed heading estimation al- gorithm. It consist of four direction paths. The proposed heading estimation algorithm uses the sensor fusion technique for estimating the direction. We compare the proposed heading estimation with azimuth based and magnetometer based approaches. The heading from all proach. The error analysis shows that the Samsung Path1 rectangular motion of pedestrian is chosen to eval- 2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. Sec http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Related Figures (18)

> has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DC A LLL 10.1109/ACCESS.2019.2891942, IEEE Access
> has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DC A LLL 10.1109/ACCESS.2019.2891942, IEEE Access
FIGURE 4: Smartphone-based position estimation. FIGURE 3: Conventional position estimation method using accelerometer and magnetometer sensors. from gyroscope are integrated with respect to time. SENSOR FUSION Sensor-fusion helps to reduce on the accumulated error from the accelerometer. The accelerometer exhibits a limited range of error when the sensor-fusion approach is applied. The expression used for calculation of the pitch
FIGURE 4: Smartphone-based position estimation. FIGURE 3: Conventional position estimation method using accelerometer and magnetometer sensors. from gyroscope are integrated with respect to time. SENSOR FUSION Sensor-fusion helps to reduce on the accumulated error from the accelerometer. The accelerometer exhibits a limited range of error when the sensor-fusion approach is applied. The expression used for calculation of the pitch
FIGURE 2: Conventional position estimation method using accelerometer and gyroscope sensors.
FIGURE 2: Conventional position estimation method using accelerometer and gyroscope sensors.
FIGURE 5: Kalman Filter algorithm. and roll angles from the accelerometer is an approxima- tion valid only in situations when the acceleration or angular rate is small. When we consider the gyroscope pitch and roll angles, the error is accumulated after numerical integration. Sensor-fusion removes the error from the accelerometer and gyroscope and it results in better performance. A Kalman filter [46] with a suitable system model that combines the complementary features of the gyroscope and accelerometer as seen in Fig. 5 is used for sensor fusion. Gyroscope quaternion values are used as the state variable, which is given by
FIGURE 5: Kalman Filter algorithm. and roll angles from the accelerometer is an approxima- tion valid only in situations when the acceleration or angular rate is small. When we consider the gyroscope pitch and roll angles, the error is accumulated after numerical integration. Sensor-fusion removes the error from the accelerometer and gyroscope and it results in better performance. A Kalman filter [46] with a suitable system model that combines the complementary features of the gyroscope and accelerometer as seen in Fig. 5 is used for sensor fusion. Gyroscope quaternion values are used as the state variable, which is given by
The three-axis gyroscope data in the device coordinate systems are (wz,Wy,w,) and the quaternion-based state update is given by 2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
The three-axis gyroscope data in the device coordinate systems are (wz,Wy,w,) and the quaternion-based state update is given by 2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data FIGURE 6: Proposed pitch-based step detector. icle has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: D¢ A LLL 10.1109/ACCESS.2019.2891942, IEEE Access block diagram of proposed pitch-based step detector is shown in Fig. 6. The pitch used as the sensor fusion inputs are from both the accelerometer and gyroscope. The maximum and minimum values of fused pitch is identified, a step is detected when a valid maximum peak (Maxima) and a valid minimum peak (Minima) are detected in a sequence in a certain interval. To avoid false step detection, a step is only detected if the pitch value is above a defined threshold value. The maxima is a maximum peak that exceeds upper thresh- old, while minima is the minimum peak lower than the ower threshold. The upper threshold is determined by summing last valid minima with a A threshold value, while the lower threshold is determined by subtracting ast valid maxima with a A threshold value, an interval time difference between maxima and minima is also determined experimentally, must be between 110ms — 400ms in order to ensure a valid step [54].
A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data FIGURE 6: Proposed pitch-based step detector. icle has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: D¢ A LLL 10.1109/ACCESS.2019.2891942, IEEE Access block diagram of proposed pitch-based step detector is shown in Fig. 6. The pitch used as the sensor fusion inputs are from both the accelerometer and gyroscope. The maximum and minimum values of fused pitch is identified, a step is detected when a valid maximum peak (Maxima) and a valid minimum peak (Minima) are detected in a sequence in a certain interval. To avoid false step detection, a step is only detected if the pitch value is above a defined threshold value. The maxima is a maximum peak that exceeds upper thresh- old, while minima is the minimum peak lower than the ower threshold. The upper threshold is determined by summing last valid minima with a A threshold value, while the lower threshold is determined by subtracting ast valid maxima with a A threshold value, an interval time difference between maxima and minima is also determined experimentally, must be between 110ms — 400ms in order to ensure a valid step [54].
A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data FIGURE 7: Example of step detection.
A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data FIGURE 7: Example of step detection.
FIGURE 8: The legs during the occurrence of a step.
FIGURE 8: The legs during the occurrence of a step.
The gyroscopic heading shows better results than the magnetometer heading. Let wz represent the body-frame x-axis gyro output, wy represent the body-frame y-axis gyro output, and w, represent the body-frame z-axis output. Then, the Euler angles are computed as [34] The other important information used for position- estimation is the heading. Heading estimation is the process of determining the direction of pedestrian move- ments. Two smartphone sensors are used to determine the heading in our proposed system. Both are used for purposes of error reduction especially during a long time span [47]. The relation between magnetic strength in the device ( hz, hy, hz) and global ( Hz, Hy, H-) coordinate system is expressed as
The gyroscopic heading shows better results than the magnetometer heading. Let wz represent the body-frame x-axis gyro output, wy represent the body-frame y-axis gyro output, and w, represent the body-frame z-axis output. Then, the Euler angles are computed as [34] The other important information used for position- estimation is the heading. Heading estimation is the process of determining the direction of pedestrian move- ments. Two smartphone sensors are used to determine the heading in our proposed system. Both are used for purposes of error reduction especially during a long time span [47]. The relation between magnetic strength in the device ( hz, hy, hz) and global ( Hz, Hy, H-) coordinate system is expressed as
A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data
A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data
FIGURE 12: Performance comparison of heading esti- mation.
FIGURE 12: Performance comparison of heading esti- mation.
FIGURE 13: Heading estimation error Distribution. A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data icle has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: Di A LLL 10.1109/ACCESS.2019.2891942, IEEE Access
FIGURE 13: Heading estimation error Distribution. A. Poulose et a/.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data icle has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: Di A LLL 10.1109/ACCESS.2019.2891942, IEEE Access
FIGURE 14: Heading error estimation.
FIGURE 14: Heading error estimation.
TABLE 2: Displacement error From the table, the proposed algorithm has less dis- placement errors compared to the conventional method. The proposed algorithm reduced the displacement error. The RMSE calculates the position errors. Fig. 16 shows the RMSE for position in rectangular motion. The maximum error from the proposed position-estimation algorithm is 2.6 m. T The maximum error from conven- tional method is 3.8 m. The proposed algorithm reduced the position error. The RSME of position in straight- ine motion of pedestrian is shown in Fig. 17. The maximum error in straight-line motion of pedestrian is
TABLE 2: Displacement error From the table, the proposed algorithm has less dis- placement errors compared to the conventional method. The proposed algorithm reduced the displacement error. The RMSE calculates the position errors. Fig. 16 shows the RMSE for position in rectangular motion. The maximum error from the proposed position-estimation algorithm is 2.6 m. T The maximum error from conven- tional method is 3.8 m. The proposed algorithm reduced the position error. The RSME of position in straight- ine motion of pedestrian is shown in Fig. 17. The maximum error in straight-line motion of pedestrian is
FIGURE 16: RMSE for position in rectangular motion.
FIGURE 16: RMSE for position in rectangular motion.
FIGURE 15: Trajectories of the ground truth, the conventional PDR approach and proposed PDR approach in different experiment scenarios. (a) Rectangular motion. (b) Straight-line motion. (c) Circular motion. A. Poulose et al.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data
FIGURE 15: Trajectories of the ground truth, the conventional PDR approach and proposed PDR approach in different experiment scenarios. (a) Rectangular motion. (b) Straight-line motion. (c) Circular motion. A. Poulose et al.: An Indoor Position-Estimation Algorithm Using Smartphone IMU Sensor Data
FIGURE 17: RMSE for position in straight-line Motion.
FIGURE 17: RMSE for position in straight-line Motion.
FIGURE 18: RMSE for position in circular motion.
FIGURE 18: RMSE for position in circular motion.
Related topics:
Activity RecognitionIndoor LocalizationBehaviour Prediction Framework
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved