CPC G01C 21/166 (2020.08) [A61B 5/112 (2013.01); G01C 22/006 (2013.01); G01C 21/206 (2013.01)] | 20 Claims |
1. A method for pedestrian inertial navigation, the method comprising:
receiving, by a processor, inertial data from an inertial measurement unit, wherein the inertial data is generated by the inertial measurement unit and the inertial data includes at least one of accelerometer data for a plurality of axes and gyroscope data for a plurality of axes, and wherein the inertial measurement unit is mounted to a foot of a user;
partitioning, by the processor, the inertial data into a plurality of partitions, wherein each partition of the inertial data is a full gait cycle;
reducing, by the processor, dimensionality of the plurality of partitions using principle component analysis, wherein reducing dimensionality includes generating an independent linear function for inertial data received from the inertial measurement unit and using a validation approach to determine an output dimension of the principle component analysis;
identifying, by the processor, a floor type of each partition, wherein the floor type of each partition is determined using an artificial neural network;
adapting, by the processor, one or more inertial navigation parameters for each partition using identified floor type, wherein a zero-velocity update threshold is adapted for a gait frequency and for floor type; and
estimating, by the processor, a navigation position of the inertial measurement unit based on adapted inertial navigation parameters.
|