User location prediction based on past measurements is extensively studied and we use one of the commonly used approach based on Hidden Markov Model. As we are specifically considering second order Hidden Markov model, which takes into consideration the direction of motion to improve the prediction performance over a first order HMM. Other methods like kalman filter can also be employed but considering the sample size due to past measurements the filter computation will be quite complex due to increase in matrix sizes. Using the HMM model algorithm, among the available sensors the one with lowest energy is selected and used in determining the location. Also it is shown in literature that the computational overhead for a second order HMM is quite negligible and can be safely ignored. At the start of algorithm a single high accurate user location information at time T0 is recorded in memory which corresponds to highly accurate user location information available. We record this information in memory as X0=
(P0,E0,T0), corresponding to position, error and time at the time of sampling. If smartphone does not detect any significant movement then we will only update time and error accumulated information in X. To determine significant movement we may invoke below logic:
1. If Wi-Fi is enabled and known Wi-Fi signatures are available from the Wi-Fi hotspot, then we measure received signal strength for each signature to estimate significant departure or arrival from any hotspot then movement variable is set accordingly.
2. Using RF ID tagging or NFC if we can compute position with desired accuracy movement variable.
3. If the movement variable is set then
(a) For small variations in position due to walking or running, the layer updates position using combination of accelerometer, compass and gyroscope.
(b) If the errors accumulated exceeds threshold error value then using the A-GPS, UMTS or GPS we compute current user location and reset the position, error and time to X0.
4. If any request from application arrives to get the user position and the accuracy requirements are met for this applications then we send the computed/stored position instead of enabling positioning sensor to compute the position.
• Following are the user options that are considered in the algorithm execution.
– Changes in desired position accuracy: Overtime users can change the position accuracy requirements.
– Sensors set preference: Applications can choose the set of sensors they want to use for their applications.
User location prediction based on past measurements is extensively studied and we use one of the commonly used approach based on Hidden Markov Model. As we are specifically considering second order Hidden Markov model, which takes into consideration the direction of motion to improve the prediction performance over a first order HMM. Other methods like kalman filter can also be employed but considering the sample size due to past measurements the filter computation will be quite complex due to increase in matrix sizes. Using the HMM model algorithm, among the available sensors the one with lowest energy is selected and used in determining the location. Also it is shown in literature that the computational overhead for a second order HMM is quite negligible and can be safely ignored. At the start of algorithm a single high accurate user location information at time T0 is recorded in memory which corresponds to highly accurate user location information available. We record this information in memory as X0=(P0,E0,T0), corresponding to position, error and time at the time of sampling. If smartphone does not detect any significant movement then we will only update time and error accumulated information in X. To determine significant movement we may invoke below logic:1. If Wi-Fi is enabled and known Wi-Fi signatures are available from the Wi-Fi hotspot, then we measure received signal strength for each signature to estimate significant departure or arrival from any hotspot then movement variable is set accordingly.2. Using RF ID tagging or NFC if we can compute position with desired accuracy movement variable.3. If the movement variable is set then(a) For small variations in position due to walking or running, the layer updates position using combination of accelerometer, compass and gyroscope. (b) If the errors accumulated exceeds threshold error value then using the A-GPS, UMTS or GPS we compute current user location and reset the position, error and time to X0.4. If any request from application arrives to get the user position and the accuracy requirements are met for this applications then we send the computed/stored position instead of enabling positioning sensor to compute the position.• Following are the user options that are considered in the algorithm execution.– Changes in desired position accuracy: Overtime users can change the position accuracy requirements. – Sensors set preference: Applications can choose the set of sensors they want to use for their applications.
การแปล กรุณารอสักครู่..
