The invention relates to the technical field of navigation and positioning, in particular to a
pedestrian dead reckoning optimization method based on improved
wave crest detection and CNN-LSTM, comprising the following steps: S1,
data acquisition: acquiring
pedestrian motion sensor data including acceleration,
angular velocity and
magnetic field intensity through an
accelerometer, a
gyroscope and a
magnetometer; s2, on the basis of an improved
wave crest detection
algorithm, real-time monitoring and analysis are carried out on the acceleration
signal by setting a dynamic threshold value, and an effective
gait cycle is extracted; s3, extracting reference features of the step length from the
pedestrian motion sensor data, training a step length prediction model of the CNN-LSTM network, and optimizing the step length prediction model; and S4, obtaining a pedestrian course angle based on
quaternion attitude calculation, realizing
gait cycle segmentation and step length prediction in combination with an improved
wave crest detection
algorithm and a CNN-LSTM model, and finally calculating the position of the pedestrian at the next moment through the course angle and the step length so as to complete continuous navigation positioning. The method can achieve the dynamic extraction and fusion of the spatial-temporal features of multiple sensors, improves the
gait detection robustness and step length
estimation precision in a complex scene, remarkably optimizes the overall positioning performance of a PDR, and solves a problem that the positioning precision is reduced because a
signal is affected by sensor
noise, behavior diversity and the like.